{"id":113344,"date":"2026-08-24T00:42:27","date_gmt":"2026-08-24T00:42:27","guid":{"rendered":"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/"},"modified":"2026-08-24T00:42:27","modified_gmt":"2026-08-24T00:42:27","slug":"harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work","status":"publish","type":"post","link":"https:\/\/youzum.net\/ja\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/","title":{"rendered":"Harvey Introduces Harvey Tenet: A Kimi K3 Base Post-Trained with Fireworks for Long-Horizon Legal Agent Work"},"content":{"rendered":"<p class=\"wp-block-paragraph\">Harvey has released <strong>Harvey Tenet<\/strong>, its first post-trained model, as a <a href=\"https:\/\/www.harvey.ai\/blog\/post-training-update-harvey-tenet\">research preview<\/a> as of today. Tenet is a Kimi K3 base post-trained with Fireworks through asynchronous reinforcement learning on long-horizon legal work. The training corpus combined synthetic data, publicly available legal data, and human expert data. Harvey states no customer data was used. Against the base K3 model, Tenet completes almost twice as many held-out tasks on Harvey\u2019s <a href=\"https:\/\/www.harvey.ai\/en-US\/blog\/introducing-harveys-legal-agent-benchmark\">Legal Agent Benchmark<\/a> (LAB) and 20% more on <a href=\"https:\/\/www.harvey.ai\/en-US\/blog\/legal-agent-benchmark-in-house-contracting\">LAB: Contracts<\/a>, raising all-pass rate by 9 and 2 percentage points respectively. Harvey reports state-of-the-art on LAB: Contracts and second place on LAB. The gains also transferred, untrained, to Mercor\u2019s <a href=\"https:\/\/www.mercor.com\/apex\/apex-agents-leaderboard\/corporate-lawyer-agent\/\">APEX Agents<\/a> and Crosby\u2019s <a href=\"https:\/\/intelligence.crosby.ai\/\">Redline Bench<\/a>. The stated goal is twofold: build frontier legal intelligence on open-weight models, and give law firms a path to own their own specialized models.<\/p>\n<h2 class=\"wp-block-heading\"><strong>Is it deployable?<\/strong><\/h2>\n<p class=\"wp-block-paragraph\"><strong>Not yet<\/strong>, <a href=\"https:\/\/www.harvey.ai\/blog\/post-training-update-harvey-tenet\">Harvey Tenet<\/a> is a research preview announced on August 20, 2026. Harvey has not published weights, a model card, or an API endpoint. The base model is open-weight; Tenet itself is Harvey\u2019s own checkpoint, and the company says the work will move \u201cfrom research to production\u201d inside Harvey\u2019s products over time. What ships today is the recipe, not the artifact.<\/p>\n<ul class=\"wp-block-list\">\n<li><strong>Company tier:<\/strong> Enterprise only. Access runs through Harvey\u2019s platform, which is sold to <a href=\"https:\/\/www.harvey.ai\/en-US\/solutions\/law-firms\">law firms<\/a>, <a href=\"https:\/\/www.harvey.ai\/en-US\/solutions\/mid-sized-firms\">mid-sized firms<\/a>, and <a href=\"https:\/\/www.harvey.ai\/en-US\/solutions\/in-house\">in-house legal teams<\/a>. A lab with an RL stack could reproduce the method; training used roughly 150 NVIDIA B300 GPUs over two months.<\/li>\n<li><strong>Industries:<\/strong> Legal services, corporate in-house legal, private equity and investment banking (M&amp;A diligence), plus regulated sectors where contract volume drives cost \u2014 insurance, financial services, healthcare, energy.<\/li>\n<li><strong>Applications:<\/strong> M&amp;A due diligence memos over datarooms, contract drafting, review and redlining, structured extraction across up to 10,000 documents, and precedent search over a firm\u2019s accumulated knowledge.<\/li>\n<\/ul>\n<h2 class=\"wp-block-heading\"><strong>What the numbers say<\/strong><\/h2>\n<p class=\"wp-block-paragraph\">Against the base K3 model, Tenet completes almost twice as many held-out tasks on Harvey\u2019s <a href=\"https:\/\/www.harvey.ai\/en-US\/blog\/introducing-harveys-legal-agent-benchmark\">Legal Agent Benchmark<\/a> (LAB) and 20% more on <a href=\"https:\/\/www.harvey.ai\/en-US\/blog\/legal-agent-benchmark-in-house-contracting\">LAB: Contracts<\/a>, lifting all-pass rate by 9 and 2 percentage points respectively. Harvey reports state-of-the-art on LAB: Contracts and second place on LAB, using base-model scores from <a href=\"https:\/\/www.vals.ai\/benchmarks\/hlab\">Vals<\/a>.<\/p>\n<p class=\"wp-block-paragraph\">The more interesting result is transfer. Tenet also improves substantially on Mercor\u2019s <a href=\"https:\/\/www.mercor.com\/apex\/apex-agents-leaderboard\/corporate-lawyer-agent\/\">APEX Agents<\/a> (corporate law) and Crosby\u2019s <a href=\"https:\/\/intelligence.crosby.ai\/\">Redline Bench<\/a> \u2014 neither seen during training \u2014 while holding performance on knowledge benchmarks including <a href=\"https:\/\/hazyresearch.stanford.edu\/legalbench\/\">LegalBench<\/a>, <a href=\"https:\/\/www.atticusprojectai.org\/cuad\/\">CUAD<\/a>, <a href=\"https:\/\/www.atticusprojectai.org\/maud\/\">MAUD<\/a>, and Scale\u2019s <a href=\"https:\/\/github.com\/scaleapi\/PRBench\">PRBench<\/a>. Agentic training did not erode textbook legal reasoning.<\/p>\n<p class=\"wp-block-paragraph\">Cost is co-optimized rather than traded away. Open weights lower price per token; reward shaping that prefers shorter trajectories at equal quality lowers tokens consumed. Harvey reports significant quality gains at stable cost.<\/p>\n<h2 class=\"wp-block-heading\"><strong>How it was trained<\/strong><\/h2>\n<p class=\"wp-block-paragraph\">Training used asynchronous reinforcement learning in sandboxed legal environments built like LAB tasks: a partner-style instruction averaging about 50 words, a client matter of key and peripheral documents, and an expert rubric of atomic pass\/fail criteria \u2014 roughly 50 per task, hundreds at the extreme. A single rollout can exceed 1,000 turns.<\/p>\n<p class=\"wp-block-paragraph\">Rollouts are graded by LLM-as-a-judge; ablations settled on Kimi 2.6. Reward combines the fraction of rubric criteria satisfied, a holistic count of legal issues solved, and an all-pass bonus. The policy is optimized with <a href=\"https:\/\/arxiv.org\/abs\/2507.18071\">GSPO<\/a> using a rank-64 LoRA over the full K3 network, eight task groups of eight rollouts per optimizer step, across ~1,750 environments and &gt;10,000 rollouts per epoch. Fireworks co-built trainer and rollout deployments at the kernel level, with token-in-token-out and router replay, to keep a large MoE numerically aligned across training and inference. <\/p>\n<p class=\"wp-block-paragraph\">\n<h2 class=\"wp-block-heading\"><strong>Three capabilities trained separately<\/strong><\/h2>\n<\/p><p class=\"wp-block-paragraph\"><strong>Harvey team also post-trained specialist models that Tenet can route to as tools or sub-agents:<\/strong><\/p>\n<ul class=\"wp-block-list\">\n<li><strong>M&amp;A diligence<\/strong>: On <a href=\"https:\/\/www.harvey.ai\/en-US\/blog\/legal-agent-bench-m-and-a-due-diligence\">LAB: Diligence<\/a>, a single task can traverse up to 80M tokens; no baseline passed more than 43.8% of criteria. With Baseten, Harvey moved to a Recursive Language Model harness where a root agent holds the dataroom in a REPL and delegates to sub-agents. A GLM-5.2 orchestrator alone reached 46.1%; post-training it in that harness via self-distillation reached 60.1%.<\/li>\n<li><strong>Review Table<\/strong>: With <a href=\"https:\/\/www.harvey.ai\/en-US\/blog\/training-frontier-review-table-models-with-applied-compute\">Applied Compute<\/a>, a post-trained GLM-5.2 improved answer quality by 3.6 points and citation quality by 12.1 points at roughly one-tenth the cost per cell, learning to abstain when a question does not apply.<\/li>\n<li><strong>Firm knowledge<\/strong>: With <a href=\"https:\/\/engram.com\/blog\/legal-agents-with-memory\">Engram<\/a>, a Qwen3.8-27B model studies ~100M tokens of client matters into 1M tokens of structured knowledge plus parametric memory. Criteria pass rate rose more than 15%, tokens in completed trajectories fell 58%, and cost per query dropped roughly 90% \u2014 190.8 intelligence-per-token versus 129.3 for the best frontier configuration.<\/li>\n<\/ul>\n<h2 class=\"wp-block-heading\"><strong>Marktechpost Independent Test Facts<\/strong><\/h2>\n<div>\n<div class=\"rc-bar\">\n<div>\n<p class=\"rc-eyebrow\">Reality Check<\/p>\n<p class=\"rc-t\">Harvey Tenet \u2014 benchmark claims, verified<\/p>\n<p class=\"rc-src\">Audited: <a href=\"https:\/\/www.harvey.ai\/blog\/post-training-update-harvey-tenet\" target=\"_blank\" rel=\"noopener\">Harvey research preview<\/a>, Aug 20, 2026 \u00b7 and the <a href=\"https:\/\/x.com\/harvey\/status\/2090454750059958440\" target=\"_blank\" rel=\"noopener\">Harvey X thread<\/a> \u00b7 Mode: default<\/p>\n<\/div>\n<div class=\"rc-chip\"><b>78<\/b><span>Inflation score<\/span><\/div>\n<\/div>\n<div class=\"rc-strip\">\n<div class=\"rc-cell\"><b>19<\/b><span>Claims<\/span><\/div>\n<div class=\"rc-cell\"><b>1<\/b><span>Verified<\/span><\/div>\n<div class=\"rc-cell\"><b>10<\/b><span>Self-reported<\/span><\/div>\n<div class=\"rc-cell\"><b>6<\/b><span>Flagged<\/span><\/div>\n<div class=\"rc-cell\"><b>2<\/b><span>Unverifiable<\/span><\/div>\n<\/div>\n<div class=\"rc-body\">\n<p class=\"rc-lead\"><b>Nothing Harvey published was contradicted.<\/b> The score is high because Tenet appears on no public leaderboard \u2014 not <a href=\"https:\/\/www.vals.ai\/benchmarks\/hlab\" target=\"_blank\" rel=\"noopener\">Vals<\/a>, not <a href=\"https:\/\/artificialanalysis.ai\/evaluations\/harvey-lab-aa\" target=\"_blank\" rel=\"noopener\">Artificial Analysis<\/a>, not <a href=\"https:\/\/www.mercor.com\/apex\/apex-agents-leaderboard\/corporate-lawyer-agent\/\" target=\"_blank\" rel=\"noopener\">Mercor<\/a>. Score formula: (8 \u00d7 6 flags) + (15 \u00d7 0 contradicted) + (3 \u00d7 10 self-reported) = 78.<\/p>\n<p class=\"rc-h\">Claim table<\/p>\n<table>\n<thead>\n<tr>\n<th>Claim<\/th>\n<th>Number<\/th>\n<th>Independent check<\/th>\n<th>Verdict<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td data-l=\"Claim\">Completes ~2\u00d7 more <a href=\"https:\/\/www.harvey.ai\/en-US\/blog\/introducing-harveys-legal-agent-benchmark\" target=\"_blank\" rel=\"noopener\">LAB<\/a> held-out tasks than Kimi K3 base<\/td>\n<td class=\"n\" data-l=\"Number\">\u2248 2\u00d7<\/td>\n<td data-l=\"Check\">Not on any public board<\/td>\n<td data-l=\"Verdict\"><span class=\"b sr\">Self-reported<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\">LAB all-pass rate lift<\/td>\n<td class=\"n\" data-l=\"Number\">+9 pts<\/td>\n<td data-l=\"Check\">Same result stated as <a href=\"https:\/\/x.com\/harvey\/status\/2090454750059958440\" target=\"_blank\" rel=\"noopener\">\u201c+82%\u201d on X<\/a><\/td>\n<td data-l=\"Verdict\"><span class=\"b fl\">Flag F1<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\"><a href=\"https:\/\/www.harvey.ai\/en-US\/blog\/legal-agent-benchmark-in-house-contracting\" target=\"_blank\" rel=\"noopener\">LAB: Contracts<\/a> all-pass lift<\/td>\n<td class=\"n\" data-l=\"Number\">+2 pts<\/td>\n<td data-l=\"Check\">No public leaderboard exists<\/td>\n<td data-l=\"Verdict\"><span class=\"b sr\">Self-reported<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\">State-of-the-art on LAB: Contracts<\/td>\n<td class=\"n\" data-l=\"Number\">SOTA<\/td>\n<td data-l=\"Check\">Benchmark owned, run and graded by Harvey<\/td>\n<td data-l=\"Verdict\"><span class=\"b fl\">Flag F2<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\">Places second on LAB<\/td>\n<td class=\"n\" data-l=\"Number\">#2<\/td>\n<td data-l=\"Check\"><a href=\"https:\/\/www.vals.ai\/home\" target=\"_blank\" rel=\"noopener\">Vals #1 is Muse Spark 1.1 at 20.00%<\/a>; Harvey-run, tool delta never quantified<\/td>\n<td data-l=\"Verdict\"><span class=\"b fl\">Flag F3<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\">Kimi K3 base on <a href=\"https:\/\/www.mercor.com\/apex\/apex-agents-leaderboard\/corporate-lawyer-agent\/\" target=\"_blank\" rel=\"noopener\">APEX Agents<\/a>, corporate law<\/td>\n<td class=\"n\" data-l=\"Number\">58.8%<\/td>\n<td data-l=\"Check\"><b>58.8% (Kimi K3 Max)<\/b> \u2014 matches exactly<\/td>\n<td data-l=\"Verdict\"><span class=\"b ok\">Verified<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\">Tenet substantially beats K3 base on APEX Agents<\/td>\n<td class=\"n\" data-l=\"Number\">\u2014<\/td>\n<td data-l=\"Check\">Harvey harness alone: 58.8% \u2192 67.5%<\/td>\n<td data-l=\"Verdict\"><span class=\"b fl\">Flag F4<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\">Beats K3 base on Crosby <a href=\"https:\/\/intelligence.crosby.ai\/benchmark\/\" target=\"_blank\" rel=\"noopener\">Redline Bench<\/a><\/td>\n<td class=\"n\" data-l=\"Number\">Not given<\/td>\n<td data-l=\"Check\">Absent from the <a href=\"https:\/\/huggingface.co\/datasets\/crosbylegal\/RedlineBench\" target=\"_blank\" rel=\"noopener\">public leaderboard<\/a><\/td>\n<td data-l=\"Verdict\"><span class=\"b sr\">Self-reported<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\"><a href=\"https:\/\/www.mercor.com\/apex\/apex-v1-leaderboard\/big-law-associate\/\" target=\"_blank\" rel=\"noopener\">APEX v1<\/a> Big Law Associate held \u2014 a <b>knowledge<\/b> benchmark, not the agentic board<\/td>\n<td class=\"n\" data-l=\"Number\">Not given<\/td>\n<td data-l=\"Check\">Blind run commissioned from Mercor; not posted to the public v1 board<\/td>\n<td data-l=\"Verdict\"><span class=\"b sr\">Self-reported<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\">Holds on <a href=\"https:\/\/hazyresearch.stanford.edu\/legalbench\/\" target=\"_blank\" rel=\"noopener\">LegalBench<\/a>, <a href=\"https:\/\/www.atticusprojectai.org\/cuad\/\" target=\"_blank\" rel=\"noopener\">CUAD<\/a>, <a href=\"https:\/\/www.atticusprojectai.org\/maud\/\" target=\"_blank\" rel=\"noopener\">MAUD<\/a><\/td>\n<td class=\"n\" data-l=\"Number\">\u201cstrong\u201d<\/td>\n<td data-l=\"Check\">Non-canonical metrics, applied to all models<\/td>\n<td data-l=\"Verdict\"><span class=\"b sr\">Self-reported<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\"><a href=\"https:\/\/github.com\/scaleapi\/PRBench\" target=\"_blank\" rel=\"noopener\">PRBench<\/a> hard subset<\/td>\n<td class=\"n\" data-l=\"Number\">36.0 \u2192 36.8%<\/td>\n<td data-l=\"Check\">Harvey calls it not statistically significant<\/td>\n<td data-l=\"Verdict\"><span class=\"b sr\">Self-reported<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\"><a href=\"https:\/\/www.harvey.ai\/en-US\/blog\/legal-agent-bench-m-and-a-due-diligence\" target=\"_blank\" rel=\"noopener\">LAB: Diligence<\/a> criteria pass rate<\/td>\n<td class=\"n\" data-l=\"Number\">43.8 \u2192 60.1%<\/td>\n<td data-l=\"Check\">No public leaderboard<\/td>\n<td data-l=\"Verdict\"><span class=\"b sr\">Self-reported<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\"><a href=\"https:\/\/www.harvey.ai\/en-US\/blog\/training-frontier-review-table-models-with-applied-compute\" target=\"_blank\" rel=\"noopener\">Review Table<\/a> cost per cell<\/td>\n<td class=\"n\" data-l=\"Number\">\u2248 1\/10<\/td>\n<td data-l=\"Check\">Baseline model never named<\/td>\n<td data-l=\"Verdict\"><span class=\"b fl\">Flag F6<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\"><a href=\"https:\/\/www.harvey.ai\/en-US\/blog\/legal-agent-bench-law-firm-knowledge\" target=\"_blank\" rel=\"noopener\">Firm Knowledge<\/a> intelligence-per-token<\/td>\n<td class=\"n\" data-l=\"Number\">190.8<\/td>\n<td data-l=\"Check\"><a href=\"https:\/\/engram.com\/blog\/legal-agents-with-memory\" target=\"_blank\" rel=\"noopener\">Engram write-up<\/a>; metric is Harvey\u2019s own<\/td>\n<td data-l=\"Verdict\"><span class=\"b sr\">Self-reported<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\">\u201cOur first post-trained <b>open-weight<\/b> model\u201d<\/td>\n<td class=\"n\" data-l=\"Number\">\u2014<\/td>\n<td data-l=\"Check\"><a href=\"https:\/\/www.techmeme.com\/260818\/p24\" target=\"_blank\" rel=\"noopener\">Business Insider: proprietary, in-house<\/a><\/td>\n<td data-l=\"Verdict\"><span class=\"b fl\">Flag F5<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\">\u201cLess than a fourth the cost of leading foundation models\u201d<\/td>\n<td class=\"n\" data-l=\"Number\">&lt; 25%<\/td>\n<td data-l=\"Check\">X only; comparators unnamed<\/td>\n<td data-l=\"Verdict\"><span class=\"b fl\">Flag F7<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\">\u2248150 NVIDIA B300 GPUs, 2 months, <a href=\"https:\/\/arxiv.org\/abs\/2507.18071\" target=\"_blank\" rel=\"noopener\">GSPO<\/a> + rank-64 LoRA<\/td>\n<td class=\"n\" data-l=\"Number\">\u2014<\/td>\n<td data-l=\"Check\">Unverifiable by construction<\/td>\n<td data-l=\"Verdict\"><span class=\"b nf\">Not checkable<\/span><\/td>\n<\/tr>\n<tr>\n<td data-l=\"Claim\">No customer data used in post-training<\/td>\n<td class=\"n\" data-l=\"Number\">\u2014<\/td>\n<td data-l=\"Check\">Unverifiable by construction<\/td>\n<td data-l=\"Verdict\"><span class=\"b nf\">Not checkable<\/span><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p class=\"rc-h\">Flags explained<\/p>\n<p class=\"rc-flag\"><b>F1 \u00b7 Denominator game<\/b>The blog reports <a href=\"https:\/\/www.harvey.ai\/blog\/post-training-update-harvey-tenet\" target=\"_blank\" rel=\"noopener\">+9 and +2 percentage points<\/a>. The <a href=\"https:\/\/x.com\/harvey\/status\/2090454750059958440\" target=\"_blank\" rel=\"noopener\">X thread<\/a> reports the same result as +82% and +22%. Both true; the social number sounds nine times larger.<\/p>\n<p class=\"rc-flag\"><b>F2 \u00b7 Self-report as fact<\/b>\u201cSOTA on LAB: Contracts\u201d is a win on Harvey\u2019s own benchmark. Harvey states there is no public leaderboard for it and that all scores are internal Harvey runs. <a href=\"https:\/\/www.lawnext.com\/2026\/05\/some-thoughts-on-harveys-launch-of-lab-an-open-source-long-horizon-benchmark-for-legal-ai-agents.html\" target=\"_blank\" rel=\"noopener\">LAB launched deliberately without a leaderboard.<\/a><\/p>\n<p class=\"rc-flag\"><b>F3 \u00b7 Settings mismatch<\/b>Harvey disclosed this plainly: Tenet ran in the standard public harness <i>plus a finish tool<\/i> carried over from training, while rival scores came from <a href=\"https:\/\/www.vals.ai\/benchmarks\/hlab\" target=\"_blank\" rel=\"noopener\">Vals<\/a>. The flag is about comparability, not concealment \u2014 Harvey never published LAB with and without the tool, so its value is unquantified. Harvey\u2019s own APEX figures show a harness change moving bare K3 by 8.7 points, and the LAB claim is a <i>rank<\/i> where Vals\u2019 leaders sit between 12% and 20%.<\/p>\n<p class=\"rc-flag\"><b>F4 \u00b7 Settings mismatch<\/b>On APEX Agents, Tenet ran in Harvey\u2019s internal bash harness while rivals used <a href=\"https:\/\/www.mercor.com\/apex\/apex-agents-leaderboard\/corporate-lawyer-agent\/\" target=\"_blank\" rel=\"noopener\">Mercor\u2019s published numbers<\/a>. Harvey discloses the harness lifts bare K3 from 58.8% to 67.5% \u2014 within 0.1 pt of leader Fable 5 at 67.4%, before any training.<\/p>\n<p class=\"rc-flag\"><b>F5 \u00b7 Framing<\/b>\u201cOpen-weight\u201d describes the Kimi K3 base, not Tenet. No weights, model card or API were published, yet multiple outlets ran headlines calling Tenet itself an open-weight release.<\/p>\n<p class=\"rc-flag\"><b>F6 \u00b7 Denominator game<\/b>\u201cRoughly one-tenth the cost per cell\u201d is measured against unnamed \u201cstrongest baselines,\u201d with no serving config, precision or hardware given for either side.<\/p>\n<p class=\"rc-flag\"><b>F7 \u00b7 Denominator game<\/b>\u201cLess than a fourth the cost of leading foundation models\u201d appears only on X. The comparators are unnamed and list price is not separated from measured token consumption.<\/p>\n<p class=\"rc-good\"><b>Credit where due<\/b>Harvey had Mercor run <a href=\"https:\/\/www.mercor.com\/apex\/apex-v1-leaderboard\/big-law-associate\/\" target=\"_blank\" rel=\"noopener\">APEX v1<\/a> blind, without disclosing runs, tasks or task-level scores back to Harvey \u2014 the strongest verification method in the post, though it evidences knowledge retention rather than agentic skill. Harvey also volunteered a null result on PRBench, documented its divergences from <a href=\"https:\/\/artificialanalysis.ai\/evaluations\/harvey-lab-aa\" target=\"_blank\" rel=\"noopener\">Artificial Analysis<\/a> and Vals, and disclosed the harness effect in F4 that undercuts its own APEX framing.<\/p>\n<\/div>\n<p class=\"rc-foot\">Reality Check by Marktechpost \u00b7 verified 2026-08-23<span>Default mode: vendor-only numbers accepted with a self-reported label. Scores change; re-verify before citing.<\/span><\/p>\n<\/div>\n<p class=\"wp-block-paragraph\">\n<h2 class=\"wp-block-heading\"><strong>Key Takeaways<\/strong><\/h2>\n<ul class=\"wp-block-list\">\n<li>Tenet is a post-trained Kimi K3 checkpoint, not a public open-weight release \u2014 no weights, no API.<\/li>\n<li>Gains transferred untrained to APEX Agents and Redline Bench, suggesting learned behavior, not benchmark fitting.<\/li>\n<li>Reward shaping on trajectory length made quality and cost improve together instead of trading off.<\/li>\n<li>The specialist stack \u2014 RLM diligence, Review Table, firm memory \u2014 is where the largest deltas landed.<\/li>\n<\/ul>\n<\/p><p class=\"wp-block-paragraph\">\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n<\/p><p class=\"wp-block-paragraph\">\n<\/p><p class=\"wp-block-paragraph\">Check out the\u00a0<strong><a href=\"https:\/\/www.harvey.ai\/blog\/post-training-update-harvey-tenet\" target=\"_blank\" rel=\"noreferrer noopener\">TECHNICAL DETAILS here<\/a><\/strong><em>.<\/em>\u00a0Also,\u00a0feel free to follow us on\u00a0<strong><a href=\"https:\/\/x.com\/intent\/follow?screen_name=marktechpost\" target=\"_blank\" rel=\"noopener\"><mark>Twitter<\/mark><\/a><\/strong>\u00a0and don\u2019t forget to join our\u00a0<strong><a href=\"https:\/\/www.reddit.com\/r\/machinelearningnews\/\" target=\"_blank\" rel=\"noopener\">150k+ML SubReddit<\/a><\/strong>\u00a0and Subscribe to\u00a0<strong><a href=\"https:\/\/magic.beehiiv.com\/v1\/f5e63dd4-5653-4f09-83e2-321a8b1ba526?email=%7B%7Bemail%7D%7D\" target=\"_blank\" rel=\"noopener\">our Newsletter<\/a><\/strong>. Wait! are you on telegram?\u00a0<strong><a href=\"https:\/\/t.me\/machinelearningresearchnews\" target=\"_blank\" rel=\"noopener\">now you can join us on telegram as well.<\/a><\/strong><\/p>\n<p class=\"wp-block-paragraph\">Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.?\u00a0<strong><a href=\"https:\/\/forms.gle\/wbash1wF6efRj8G58\" target=\"_blank\" rel=\"noopener\"><mark>Connect with us<\/mark><\/a><\/strong><\/p>\n<p>The post <a href=\"https:\/\/www.marktechpost.com\/2026\/08\/23\/harvey-tenet-post-trained-kimi-k3-legal-agent-model\/\">Harvey Introduces Harvey Tenet: A Kimi K3 Base Post-Trained with Fireworks for Long-Horizon Legal Agent Work<\/a> appeared first on <a href=\"https:\/\/www.marktechpost.com\/\">MarkTechPost<\/a>.<\/p>","protected":false},"excerpt":{"rendered":"<p>Harvey has released Harvey Tenet, its first post-trained model, as a research preview as of today. Tenet is a Kimi K3 base post-trained with Fireworks through asynchronous reinforcement learning on long-horizon legal work. The training corpus combined synthetic data, publicly available legal data, and human expert data. Harvey states no customer data was used. Against the base K3 model, Tenet completes almost twice as many held-out tasks on Harvey\u2019s Legal Agent Benchmark (LAB) and 20% more on LAB: Contracts, raising all-pass rate by 9 and 2 percentage points respectively. Harvey reports state-of-the-art on LAB: Contracts and second place on LAB. The gains also transferred, untrained, to Mercor\u2019s APEX Agents and Crosby\u2019s Redline Bench. The stated goal is twofold: build frontier legal intelligence on open-weight models, and give law firms a path to own their own specialized models. Is it deployable? Not yet, Harvey Tenet is a research preview announced on August 20, 2026. Harvey has not published weights, a model card, or an API endpoint. The base model is open-weight; Tenet itself is Harvey\u2019s own checkpoint, and the company says the work will move \u201cfrom research to production\u201d inside Harvey\u2019s products over time. What ships today is the recipe, not the artifact. Company tier: Enterprise only. Access runs through Harvey\u2019s platform, which is sold to law firms, mid-sized firms, and in-house legal teams. A lab with an RL stack could reproduce the method; training used roughly 150 NVIDIA B300 GPUs over two months. Industries: Legal services, corporate in-house legal, private equity and investment banking (M&amp;A diligence), plus regulated sectors where contract volume drives cost \u2014 insurance, financial services, healthcare, energy. Applications: M&amp;A due diligence memos over datarooms, contract drafting, review and redlining, structured extraction across up to 10,000 documents, and precedent search over a firm\u2019s accumulated knowledge. What the numbers say Against the base K3 model, Tenet completes almost twice as many held-out tasks on Harvey\u2019s Legal Agent Benchmark (LAB) and 20% more on LAB: Contracts, lifting all-pass rate by 9 and 2 percentage points respectively. Harvey reports state-of-the-art on LAB: Contracts and second place on LAB, using base-model scores from Vals. The more interesting result is transfer. Tenet also improves substantially on Mercor\u2019s APEX Agents (corporate law) and Crosby\u2019s Redline Bench \u2014 neither seen during training \u2014 while holding performance on knowledge benchmarks including LegalBench, CUAD, MAUD, and Scale\u2019s PRBench. Agentic training did not erode textbook legal reasoning. Cost is co-optimized rather than traded away. Open weights lower price per token; reward shaping that prefers shorter trajectories at equal quality lowers tokens consumed. Harvey reports significant quality gains at stable cost. How it was trained Training used asynchronous reinforcement learning in sandboxed legal environments built like LAB tasks: a partner-style instruction averaging about 50 words, a client matter of key and peripheral documents, and an expert rubric of atomic pass\/fail criteria \u2014 roughly 50 per task, hundreds at the extreme. A single rollout can exceed 1,000 turns. Rollouts are graded by LLM-as-a-judge; ablations settled on Kimi 2.6. Reward combines the fraction of rubric criteria satisfied, a holistic count of legal issues solved, and an all-pass bonus. The policy is optimized with GSPO using a rank-64 LoRA over the full K3 network, eight task groups of eight rollouts per optimizer step, across ~1,750 environments and &gt;10,000 rollouts per epoch. Fireworks co-built trainer and rollout deployments at the kernel level, with token-in-token-out and router replay, to keep a large MoE numerically aligned across training and inference. Three capabilities trained separately Harvey team also post-trained specialist models that Tenet can route to as tools or sub-agents: M&amp;A diligence: On LAB: Diligence, a single task can traverse up to 80M tokens; no baseline passed more than 43.8% of criteria. With Baseten, Harvey moved to a Recursive Language Model harness where a root agent holds the dataroom in a REPL and delegates to sub-agents. A GLM-5.2 orchestrator alone reached 46.1%; post-training it in that harness via self-distillation reached 60.1%. Review Table: With Applied Compute, a post-trained GLM-5.2 improved answer quality by 3.6 points and citation quality by 12.1 points at roughly one-tenth the cost per cell, learning to abstain when a question does not apply. Firm knowledge: With Engram, a Qwen3.8-27B model studies ~100M tokens of client matters into 1M tokens of structured knowledge plus parametric memory. Criteria pass rate rose more than 15%, tokens in completed trajectories fell 58%, and cost per query dropped roughly 90% \u2014 190.8 intelligence-per-token versus 129.3 for the best frontier configuration. Marktechpost Independent Test Facts Reality Check Harvey Tenet \u2014 benchmark claims, verified Audited: Harvey research preview, Aug 20, 2026 \u00b7 and the Harvey X thread \u00b7 Mode: default 78Inflation score 19Claims 1Verified 10Self-reported 6Flagged 2Unverifiable Nothing Harvey published was contradicted. The score is high because Tenet appears on no public leaderboard \u2014 not Vals, not Artificial Analysis, not Mercor. Score formula: (8 \u00d7 6 flags) + (15 \u00d7 0 contradicted) + (3 \u00d7 10 self-reported) = 78. Claim table Claim Number Independent check Verdict Completes ~2\u00d7 more LAB held-out tasks than Kimi K3 base \u2248 2\u00d7 Not on any public board Self-reported LAB all-pass rate lift +9 pts Same result stated as \u201c+82%\u201d on X Flag F1 LAB: Contracts all-pass lift +2 pts No public leaderboard exists Self-reported State-of-the-art on LAB: Contracts SOTA Benchmark owned, run and graded by Harvey Flag F2 Places second on LAB #2 Vals #1 is Muse Spark 1.1 at 20.00%; Harvey-run, tool delta never quantified Flag F3 Kimi K3 base on APEX Agents, corporate law 58.8% 58.8% (Kimi K3 Max) \u2014 matches exactly Verified Tenet substantially beats K3 base on APEX Agents \u2014 Harvey harness alone: 58.8% \u2192 67.5% Flag F4 Beats K3 base on Crosby Redline Bench Not given Absent from the public leaderboard Self-reported APEX v1 Big Law Associate held \u2014 a knowledge benchmark, not the agentic board Not given Blind run commissioned from Mercor; not posted to the public v1 board Self-reported Holds on LegalBench, CUAD, MAUD \u201cstrong\u201d Non-canonical metrics, applied to all models Self-reported PRBench hard subset<\/p>","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"pmpro_default_level":"","site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"_pvb_checkbox_block_on_post":false,"footnotes":""},"categories":[52,5,7,1],"tags":[],"class_list":["post-113344","post","type-post","status-publish","format-standard","hentry","category-ai-club","category-committee","category-news","category-uncategorized","pmpro-has-access"],"acf":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v25.3 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Harvey Introduces Harvey Tenet: A Kimi K3 Base Post-Trained with Fireworks for Long-Horizon Legal Agent Work - YouZum<\/title>\n<meta name=\"description\" content=\"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/youzum.net\/ja\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/\" \/>\n<meta property=\"og:locale\" content=\"ja_JP\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Harvey Introduces Harvey Tenet: A Kimi K3 Base Post-Trained with Fireworks for Long-Horizon Legal Agent Work - YouZum\" \/>\n<meta property=\"og:description\" content=\"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19\" \/>\n<meta property=\"og:url\" content=\"https:\/\/youzum.net\/ja\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/\" \/>\n<meta property=\"og:site_name\" content=\"YouZum\" \/>\n<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/DroneAssociationTH\/\" \/>\n<meta property=\"article:published_time\" content=\"2026-08-24T00:42:27+00:00\" \/>\n<meta name=\"author\" content=\"admin NU\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u57f7\u7b46\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin NU\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u63a8\u5b9a\u8aad\u307f\u53d6\u308a\u6642\u9593\" \/>\n\t<meta name=\"twitter:data2\" content=\"8\u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/#article\",\"isPartOf\":{\"@id\":\"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/\"},\"author\":{\"name\":\"admin NU\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c\"},\"headline\":\"Harvey Introduces Harvey Tenet: A Kimi K3 Base Post-Trained with Fireworks for Long-Horizon Legal Agent Work\",\"datePublished\":\"2026-08-24T00:42:27+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/\"},\"wordCount\":1559,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#organization\"},\"articleSection\":[\"AI\",\"Committee\",\"News\",\"Uncategorized\"],\"inLanguage\":\"ja\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/\",\"url\":\"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/\",\"name\":\"Harvey Introduces Harvey Tenet: A Kimi K3 Base Post-Trained with Fireworks for Long-Horizon Legal Agent Work - YouZum\",\"isPartOf\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#website\"},\"datePublished\":\"2026-08-24T00:42:27+00:00\",\"description\":\"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19\",\"breadcrumb\":{\"@id\":\"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/#breadcrumb\"},\"inLanguage\":\"ja\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\/\/youzum.net\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Harvey Introduces Harvey Tenet: A Kimi K3 Base Post-Trained with Fireworks for Long-Horizon Legal Agent Work\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/yousum.gpucore.co\/#website\",\"url\":\"https:\/\/yousum.gpucore.co\/\",\"name\":\"YouSum\",\"description\":\"\",\"publisher\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/yousum.gpucore.co\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"ja\"},{\"@type\":\"Organization\",\"@id\":\"https:\/\/yousum.gpucore.co\/#organization\",\"name\":\"Drone Association Thailand\",\"url\":\"https:\/\/yousum.gpucore.co\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"ja\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/\",\"url\":\"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png\",\"contentUrl\":\"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png\",\"width\":300,\"height\":300,\"caption\":\"Drone Association Thailand\"},\"image\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/\"},\"sameAs\":[\"https:\/\/www.facebook.com\/DroneAssociationTH\/\"]},{\"@type\":\"Person\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c\",\"name\":\"admin NU\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"ja\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png\",\"contentUrl\":\"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png\",\"caption\":\"admin NU\"},\"url\":\"https:\/\/youzum.net\/ja\/members\/adminnu\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Harvey Introduces Harvey Tenet: A Kimi K3 Base Post-Trained with Fireworks for Long-Horizon Legal Agent Work - YouZum","description":"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/youzum.net\/ja\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/","og_locale":"ja_JP","og_type":"article","og_title":"Harvey Introduces Harvey Tenet: A Kimi K3 Base Post-Trained with Fireworks for Long-Horizon Legal Agent Work - YouZum","og_description":"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19","og_url":"https:\/\/youzum.net\/ja\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/","og_site_name":"YouZum","article_publisher":"https:\/\/www.facebook.com\/DroneAssociationTH\/","article_published_time":"2026-08-24T00:42:27+00:00","author":"admin NU","twitter_card":"summary_large_image","twitter_misc":{"\u57f7\u7b46\u8005":"admin NU","\u63a8\u5b9a\u8aad\u307f\u53d6\u308a\u6642\u9593":"8\u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/#article","isPartOf":{"@id":"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/"},"author":{"name":"admin NU","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c"},"headline":"Harvey Introduces Harvey Tenet: A Kimi K3 Base Post-Trained with Fireworks for Long-Horizon Legal Agent Work","datePublished":"2026-08-24T00:42:27+00:00","mainEntityOfPage":{"@id":"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/"},"wordCount":1559,"commentCount":0,"publisher":{"@id":"https:\/\/yousum.gpucore.co\/#organization"},"articleSection":["AI","Committee","News","Uncategorized"],"inLanguage":"ja","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/#respond"]}]},{"@type":"WebPage","@id":"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/","url":"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/","name":"Harvey Introduces Harvey Tenet: A Kimi K3 Base Post-Trained with Fireworks for Long-Horizon Legal Agent Work - YouZum","isPartOf":{"@id":"https:\/\/yousum.gpucore.co\/#website"},"datePublished":"2026-08-24T00:42:27+00:00","description":"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19","breadcrumb":{"@id":"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/#breadcrumb"},"inLanguage":"ja","potentialAction":[{"@type":"ReadAction","target":["https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/youzum.net\/harvey-introduces-harvey-tenet-a-kimi-k3-base-post-trained-with-fireworks-for-long-horizon-legal-agent-work\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/youzum.net\/"},{"@type":"ListItem","position":2,"name":"Harvey Introduces Harvey Tenet: A Kimi K3 Base Post-Trained with Fireworks for Long-Horizon Legal Agent Work"}]},{"@type":"WebSite","@id":"https:\/\/yousum.gpucore.co\/#website","url":"https:\/\/yousum.gpucore.co\/","name":"YouSum","description":"","publisher":{"@id":"https:\/\/yousum.gpucore.co\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/yousum.gpucore.co\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"ja"},{"@type":"Organization","@id":"https:\/\/yousum.gpucore.co\/#organization","name":"Drone Association Thailand","url":"https:\/\/yousum.gpucore.co\/","logo":{"@type":"ImageObject","inLanguage":"ja","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/","url":"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png","contentUrl":"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png","width":300,"height":300,"caption":"Drone Association Thailand"},"image":{"@id":"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.facebook.com\/DroneAssociationTH\/"]},{"@type":"Person","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c","name":"admin NU","image":{"@type":"ImageObject","inLanguage":"ja","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/person\/image\/","url":"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png","contentUrl":"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png","caption":"admin NU"},"url":"https:\/\/youzum.net\/ja\/members\/adminnu\/"}]}},"rttpg_featured_image_url":null,"rttpg_author":{"display_name":"admin NU","author_link":"https:\/\/youzum.net\/ja\/members\/adminnu\/"},"rttpg_comment":0,"rttpg_category":"<a href=\"https:\/\/youzum.net\/ja\/category\/ai-club\/\" rel=\"category tag\">AI<\/a> <a href=\"https:\/\/youzum.net\/ja\/category\/committee\/\" rel=\"category tag\">Committee<\/a> <a href=\"https:\/\/youzum.net\/ja\/category\/news\/\" rel=\"category tag\">News<\/a> <a href=\"https:\/\/youzum.net\/ja\/category\/uncategorized\/\" rel=\"category tag\">Uncategorized<\/a>","rttpg_excerpt":"Harvey has released Harvey Tenet, its first post-trained model, as a research preview as of today. Tenet is a Kimi K3 base post-trained with Fireworks through asynchronous reinforcement learning on long-horizon legal work. The training corpus combined synthetic data, publicly available legal data, and human expert data. Harvey states no customer data was used. Against&hellip;","_links":{"self":[{"href":"https:\/\/youzum.net\/ja\/wp-json\/wp\/v2\/posts\/113344","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/youzum.net\/ja\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/youzum.net\/ja\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/youzum.net\/ja\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/youzum.net\/ja\/wp-json\/wp\/v2\/comments?post=113344"}],"version-history":[{"count":0,"href":"https:\/\/youzum.net\/ja\/wp-json\/wp\/v2\/posts\/113344\/revisions"}],"wp:attachment":[{"href":"https:\/\/youzum.net\/ja\/wp-json\/wp\/v2\/media?parent=113344"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/youzum.net\/ja\/wp-json\/wp\/v2\/categories?post=113344"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/youzum.net\/ja\/wp-json\/wp\/v2\/tags?post=113344"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}