1d:I[70616,["/_next/static/chunks/0kcf5dt6yxdox.js","/_next/static/chunks/1su5n1uv1ti4s.js","/_next/static/chunks/0gvke1wv6udfg.js","/_next/static/chunks/1dfrle3jgmlzh.js","/_next/static/chunks/162lgvduyr-fa.js","/_next/static/chunks/2gcnnopam8e9w.js","/_next/static/chunks/2g60iab5a1pkg.js"],"Breadcrumb"] 1e:I[90762,["/_next/static/chunks/0kcf5dt6yxdox.js","/_next/static/chunks/1su5n1uv1ti4s.js","/_next/static/chunks/0gvke1wv6udfg.js","/_next/static/chunks/1dfrle3jgmlzh.js","/_next/static/chunks/162lgvduyr-fa.js","/_next/static/chunks/2gcnnopam8e9w.js","/_next/static/chunks/2g60iab5a1pkg.js"],"ArticleGrid"] 1f:T1582,

Cognition released SWE-2 on Thursday, a coding model the company says lands within one point of Fable 5.1 on its own FrontierCode benchmark while costing 64% less to run. The more interesting detail is where the model came from: SWE-2 is post-trained from Kimi K3, the 2.8-trillion-parameter open-weight model from Moonshot AI, which makes it a frontier-adjacent product built on a Chinese base. The launch post reached the front page of Hacker News with 243 points.

Key takeaways

SWE-2 scores 50.0% on FrontierCode 1.1 Main against Fable 5.1's 50.9%, at roughly a third of the price of the frontier models it trails.
Cognition post-trained the model from Moonshot AI's open-weight Kimi K3 and says its reinforcement learning added 5–6 points across many benchmarks.
SWE-2 medium beats the older SWE-1.7 while using 58% fewer turns and costing 81% less per task.

How SWE-2 ranks against the frontier

On FrontierCode 1.1 Main, Cognition reports 50.0% for SWE-2, ahead of Kimi K3 at 44.2%, Grok 4.6 at 48.0% and GPT-5.6 Sol at 47.5%, and behind only Fable 5.1 at 50.9% and GPT-6 Astra at 53.3%. On DeepSWE 1.1 the model hits 73.0%, roughly a point off Astra. Terminal-Bench 2.1 is the clean win, where SWE-2 tops the table at 92.8%.

The picture is less flattering on Terminal-Bench 4, a harder suite, where SWE-2 manages 27.3% against 57.9% for Astra and 55.8% for Fable 5.1. That gap is the honest read on what a cheaper model still gives up: on the hardest long-horizon agentic tasks, the frontier remains well ahead.

Training every effort level in one run

The technical claim behind the pricing is a change to how Cognition runs reinforcement learning. Instead of training separate low, medium and high reasoning modes, the team trains all effort levels inside a single RL run using a cost-penalized reward — success minus a per-effort-level coefficient multiplied by rollout cost, where that coefficient is tuned to the local slope of the base model's cost-performance curve.

Cognition argues the penalty has to be linear in cost, since only a linear term gives the same answer whether it is applied before or after averaging. Setting the coefficient to match the frontier's tangent, rather than treating it as a hyperparameter to sweep, is what keeps the optimizer from quietly collapsing a high-effort mode into a cheaper medium-effort one.

Serving changes carried the rest. The team batched nearby prefill requests to lift throughput 10–20%, trained a fresh speculative decoding draft model that accepted 15% longer sequences, and moved to NVFP4 and FP8 kernels with quantization-aware training. Despite a base model with nearly three times the parameters of its predecessor, Cognition reports lower train-inference divergence than SWE-1.7.

Fewer detours, earlier edits

Users complained that SWE-1.7 over-explored simple tasks, reading far more of a codebase than a small fix warranted. SWE-2 makes its first real edit after a median of 18 steps on FrontierCode 1.1 Main, compared with 48 for SWE-1.7. Cognition frames this as focused exploration rather than laziness: the model is better at judging which parts of a repository matter.

The company also says SWE-2 re-derives its conclusions when challenged instead of folding, and runs artifacts to gather evidence rather than trusting prose. SWE-2 is live now in Devin Desktop and CLI, with rollout to Devin Web and Fusion underway.

What it means for coding agent pricing

Cognition's bet, consistent with the pitch behind its $2 billion raise at a $48 billion valuation, is that most engineering work does not need the most expensive model in the market. Taking a strong open-weight base and spending the compute on task-specific RL is now a credible route to within a point or two of the frontier, and it is a route that does not require pretraining a multi-trillion-parameter model from scratch.

Cognition also published trustworthiness results, reporting that SWE-2 answered 98.0% of politically sensitive prompts without adopting official Chinese state positions, including 95.2% when asked in Simplified Chinese. That evaluation matters precisely because the base model is Moonshot AI's, and it is the kind of number buyers will now expect from anyone shipping a derivative of an open Chinese model.

FAQ

Is SWE-2 open source?

No. SWE-2 is a proprietary model served through Cognition's Devin products. Its base model, Kimi K3 from Moonshot AI, is open weight, but Cognition has not released the post-trained weights.

How much cheaper is SWE-2 than comparable models?

Cognition reports SWE-2 costs 64% less than Fable 5.1 while scoring within one point of it on FrontierCode 1.1 Main, and roughly a quarter the cost of GPT-6 Astra. Against Cognition's own SWE-1.7, SWE-2 medium costs 81% less per task on the same benchmark.

Where can I use SWE-2 today?

SWE-2 is available immediately in Devin Desktop and the Devin CLI. Cognition says it is also rolling the model out across Devin Web and Fusion.

a:["$","div",null,{"className":"mx-auto max-w-7xl px-4 py-8 sm:px-6","children":[["$","script",null,{"type":"application/ld+json","dangerouslySetInnerHTML":{"__html":"{\"@context\":\"https://schema.org\",\"@type\":\"BreadcrumbList\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https://ainewsway.com\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"#ai-reasoning\",\"item\":\"https://ainewsway.com/tag/ai-reasoning\"}]}"}}],["$","script",null,{"type":"application/ld+json","dangerouslySetInnerHTML":{"__html":"{\"@context\":\"https://schema.org\",\"@type\":\"ItemList\",\"@id\":\"https://ainewsway.com/tag/ai-reasoning#itemlist\",\"name\":\"#ai-reasoning Articles\",\"numberOfItems\":1,\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"url\":\"https://ainewsway.com/articles/cognitions-swe-2-gets-within-a-point-of-the-frontier-for-64-less-mtw9ji0o\",\"name\":\"Cognition's SWE-2 Gets Within a Point of the Frontier for 64% Less\"}]}"}}],["$","$L1d",null,{"items":[{"label":"# ai-reasoning"}]}],["$","h1",null,{"className":"mb-8 text-3xl font-bold text-zinc-900 dark:text-zinc-100","children":["#","ai-reasoning"]}],[["$","$L1e",null,{"articles":[{"id":"4fec7a46-9019-4e98-b602-6c27fc917255","slug":"cognitions-swe-2-gets-within-a-point-of-the-frontier-for-64-less-mtw9ji0o","title":"Cognition's SWE-2 Gets Within a Point of the Frontier for 64% Less","subtitle":"The coding model is post-trained from Moonshot AI's open-weight Kimi K3, and trains every reasoning-effort level in a single RL run","content":"$1f","excerpt":"Cognition says SWE-2 scores 50.0% on FrontierCode 1.1 Main, a point behind Fable 5.1, at 64% lower cost — built on an open Chinese base model.","coverImage":"https://cdn.ainewsway.com/uploads/2026/09/e9c473f8-110b-465d-b430-f34bbcc8f988.webp","coverImageCaption":"Source code on a developer's screen — Cognition says SWE-2 starts editing after a median of 18 steps, against 48 for its previous coding model.","authorId":"6773f67d-d7fa-4b19-8f38-384f10ac761b","status":"PUBLISHED","publishedAt":"2026-09-11T01:14:44.082Z","viewCount":1,"likeCount":0,"readingTime":5,"isFeatured":false,"isBreakingNews":false,"metaTitle":"Cognition SWE-2: Near-Frontier Coding at 64% Less Cost","metaDescription":"Cognition's SWE-2 scores 50.0% on FrontierCode 1.1 Main, one point behind Fable 5.1 at 64% lower cost, post-trained from the open-weight Kimi K3.","metaKeywords":["Cognition SWE-2","AI coding model","FrontierCode benchmark","Kimi K3 open weights","Devin coding agent","reinforcement learning post-training","coding agent pricing"],"ogImage":null,"aiSummary":"Cognition launched SWE-2 on September 10, 2026, a coding model scoring 50.0% on FrontierCode 1.1 Main against Fable 5.1's 50.9% at 64% lower cost. It is post-trained from Moonshot AI's open-weight Kimi K3 using a reinforcement learning run that tunes all reasoning-effort levels at once. The release shows that RL on a strong open base can reach near-frontier coding quality without pretraining from scratch.","aiTags":[],"sentiment":null,"articleType":"NEWS","videoEmbed":null,"videoType":null,"seoScore":85,"isAutoGenerated":true,"isBackfill":false,"pipelineStage":null,"copyrightScore":null,"similarityScore":null,"matchedSourceId":null,"createdAt":"2026-09-11T01:14:01.849Z","updatedAt":"2026-09-15T11:32:43.011Z","contentUpdatedAt":"2026-09-15T11:32:43.010Z","deletedAt":null,"author":{"id":"6773f67d-d7fa-4b19-8f38-384f10ac761b","name":"Seung Jung","username":"Seung Jung","avatar":null},"categories":[{"id":"32f2303b-b593-473f-bf9c-6a99acbe6ce1","name":"Developer Tools","slug":"developer-tools"}],"tags":[{"id":"18cdd61e-c75a-4aff-843e-db288dddc418","name":"Coding Agents","slug":"coding-agents"},{"id":"58e48f52-38f7-4fdc-87ce-a46ab1db7c60","name":"AI Benchmarks","slug":"ai-benchmarks"},{"id":"b04c30ce-c1e3-4d6d-8cbc-a79e2862a3fa","name":"AI Code Generation","slug":"ai-code-generation"},{"id":"0aad1028-055f-4322-b7ce-a7ce5b46a44c","name":"Open Weights","slug":"open-weights"},{"id":"e909dcd8-f32f-49b1-9b71-2f5bc89d99ee","name":"LLM Pricing","slug":"llm-pricing"},{"id":"82eaa6fd-55cf-4e99-b35e-739722b7d829","name":"AI Reasoning","slug":"ai-reasoning"}],"translations":[{"id":"898d5fdc-b0b8-45ab-b085-9df45068e43c","language":"ko","title":"Cognition SWE-2, 프런티어에 1점 차로 따라붙고 비용은 64% 낮췄다","subtitle":"Moonshot AI의 오픈 웨이트 Kimi K3를 후속 학습했고, 모든 추론 강도를 한 번의 RL 실행으로 훈련한다","excerpt":"Cognition은 SWE-2가 FrontierCode 1.1 Main에서 50.0%로 Fable 5.1에 1점 뒤지면서 비용은 64% 낮다고 밝혔다. 베이스는 중국산 오픈 모델이다.","metaTitle":"Cognition SWE-2, 64% 싼 값에 프런티어급 코딩","metaDescription":"Cognition SWE-2가 FrontierCode 1.1 Main에서 50.0%를 기록해 Fable 5.1에 1점 차로 근접했다. 비용은 64% 낮고 베이스는 오픈 웨이트 Kimi K3다."},{"id":"aa609aba-f70d-4e0e-a52c-cb7163e67534","language":"ja","title":"Cognition SWE-2、フロンティアに1ポイント差で64%安い","subtitle":"Moonshot AIのオープンウェイトKimi K3を事後学習し、すべての推論強度を一度のRL実行で訓練する","excerpt":"CognitionはSWE-2がFrontierCode 1.1 Mainで50.0%を記録し、Fable 5.1に1ポイント差でコストは64%低いと発表した。ベースは中国発のオープンモデルだ。","metaTitle":"Cognition SWE-2、64%安いフロンティア級コーディング","metaDescription":"Cognition SWE-2がFrontierCode 1.1 Mainで50.0%、Fable 5.1に1ポイント差でコストは64%低い。ベースはオープンウェイトのKimi K3。"},{"id":"6c29f8cc-7118-43cb-bd8a-838cdc32a5b0","language":"zh","title":"Cognition SWE-2逼近前沿僅差一分，成本低64%","subtitle":"該程式設計模型由Moonshot AI的開源權重Kimi K3後訓練而來，並在一次強化學習中訓練全部推理強度","excerpt":"Cognition稱SWE-2在FrontierCode 1.1 Main上得分50.0%，僅落後Fable 5.1一分，成本卻低64%，基座是一款中國開源模型。","metaTitle":"Cognition SWE-2：成本低64%的準前沿程式設計模型","metaDescription":"Cognition SWE-2在FrontierCode 1.1 Main上取得50.0%，僅落後Fable 5.1一分而成本低64%，由開源權重Kimi K3後訓練而來。"},{"id":"0a0a4140-6d1f-494f-8765-8b239473a047","language":"es","title":"SWE-2 de Cognition queda a un punto de la frontera por un 64% menos","subtitle":"El modelo de programación está posentrenado sobre el Kimi K3 de pesos abiertos de Moonshot AI y entrena todos los niveles de esfuerzo en una sola corrida de RL","excerpt":"Cognition afirma que SWE-2 obtiene 50,0% en FrontierCode 1.1 Main, un punto por detrás de Fable 5.1, con un coste 64% menor y sobre una base china abierta.","metaTitle":"Cognition SWE-2: casi frontera con un 64% menos de coste","metaDescription":"SWE-2 de Cognition logra 50,0% en FrontierCode 1.1 Main, a un punto de Fable 5.1 y con un 64% menos de coste, posentrenado sobre el abierto Kimi K3."}],"celebrities":[{"id":"b8bff3fa-d9d4-4037-95df-e725a8b9749a","nameKo":"코그니션","nameEn":"Cognition","slug":"cognition","profileImage":null},{"id":"497c3ea2-a3a6-464f-930e-acb3c26831d5","nameKo":"문샷 AI","nameEn":"Moonshot AI","slug":"moonshot-ai","profileImage":null}]}]}],"$L20"]]}]

#ai-reasoning

Cognition SWE-2、フロンティアに1ポイント差で64%安い

ニュースレター購読

#ai-reasoning

Cognition SWE-2、フロンティアに1ポイント差で64%安い