SWE-bench Verified final standings and the newest models, October 2026
Last verified against vals.ai and maker sources. Ranked scores last changed . vals.ai archived its SWE-bench Verified board on , so the independent scores on this board are its final standings.
| # | Model | SWE-bench Verified | SWE-bench Pro | Input | Best for |
|---|---|---|---|---|---|
| 1 | Claude Opus 5 ↗Anthropic | 97.0%independent (vals.ai) | — | $5/1M | The highest independently measured coding score on the board at 97.0%, at half the price of Fable 5. Strongest on short and medium tasks, though GPT-5.6 Sol still edges it on multi-hour work.Independent (vals.ai, observed Jul 25 2026, mini-swe-agent bash-only harness): SWE-bench Verified 97.00% ±0.76, the highest score on the board and… full note → |
| 2 | GPT-5.6 Sol ↗OpenAI | 96.2%independent (vals.ai) | — | $5/1M | The strongest model on long tasks: 98% on the 1-to-4-hour tier, ahead of Claude Opus 5, and the second-highest overall score.Independent (vals.ai, Jul 14 2026, mini-swe-agent bash-only harness): SWE-bench Verified 96.20% ±0.86 — now the second-highest score on the board… full note → |
| 3 | Grok 4.6 ↗SpaceXAI (xAI) | 95.6%independent (vals.ai) | — | $2/1M | A post-training refresh of Grok 4.5 that lands in the top group on the neutral harness, at a third the price of the models around it.Independent (vals.ai, benchmark updated 2026-08-12, mini-swe-agent bash-only harness): SWE-bench Verified 95.60% ±0.92. Entered ranked 2026-08-13… full note → |
| 4 | GPT-5.6 Terra ↗OpenAI | 95.4%independent (vals.ai) | — | $2/1M | A mid-tier price with top-tier coding after vals.ai revised its score upward by 20 points. At $2/$12 per 1M it now scores within error bars of Sol, which costs 2.5x more on output.Independent (vals.ai, re-evaluated, mini-swe-agent bash-only harness): SWE-bench Verified 95.4% ±0.94, 5th of 86 systems. CORRECTION, 2026-08-20… full note → |
| 5 | GLM-5.3 ↗Z.ai (Zhipu AI) | 95.4%independent (vals.ai) | — | — | The highest-scoring model on this board whose maker has promised open weights, and the cheapest route to a 95%-plus measured score at $0.34 per test.Independent (vals.ai, 2026-08-20 sweep, mini-swe-agent bash-only harness): SWE-bench Verified 95.4% ±0.94, 6th of 86 systems. Left the verifying… full note → |
| 6 | Claude Fable 5 ↗Anthropic | 95.0%independent (vals.ai) | 80.3% | $10/1M | Mythos-class flagship for long-horizon agentic runs: the model to reach for when a task spans hours and hundreds of tool calls and has to actually finish.Independent (vals.ai, Jul 14 2026, mini-swe-agent bash-only harness): SWE-bench Verified 95.00% ±0.98. Held the top score until GPT-5.6 Sol was… full note → |
| 7 | Kimi K3 open ↗Moonshot AI | 93.4%independent (vals.ai) | — | $3/1M | The highest-scoring downloadable coding model we track, and by far the cheapest way to buy a 90%+ result at $3 per 1M input. Weights shipped Jul 27, 2026 as 96 shards totalling 1.56 TB, 4-bit MXFP4 only: roughly 1,454 GiB, which fits one eight-card 192GB node. No full-precision base checkpoint was published.Independent (vals.ai, mini-swe-agent bash-only harness): SWE-bench Verified 93.40% ±1.11. Verified Jul 18, 2026 — vals.ai had not run K3 at our Jul… full note → |
| 8 | GPT-5.6 Luna ↗OpenAI | 93.0%independent (vals.ai) | — | $0.20/1M | The cost outlier of the leaders: within striking distance of the top scores while costing a fraction per solved task, and after an 80% price cut on Jul 30, 2026 it is now the cheapest ranked model on the board per token.Independent (vals.ai, eval listed Jul 17 2026, mini-swe-agent bash-only harness): SWE-bench Verified 93.00% ±1.14. Added Jul 21, 2026 — this row was… full note → |
| 9 | Ember-1 ↗Fireworks AI (built on Moonshot Kimi K3) | 92.2%vendor-reported | — | — | Kimi K3 post-trained by Fireworks to reason in fewer tokens: the pick if your agent bill is dominated by long reasoning traces and you can live with a research-preview endpoint.Vendor-reported (Fireworks, Sept 23 2026, Fireworks harness): SWE-bench Verified 92.2% using 15.5% fewer tokens than Kimi K3; Terminal Bench 2.1… full note → |
| 10 | DeepSeek V4 Flash open ↗DeepSeek | 88.8%independent (vals.ai) | — | $0.14/1M | The cheapest model making a near-frontier agentic claim: 284B MoE with 13B active, 1M context, and native Responses API plus Codex support at $0.28 per 1M output.Independent (vals.ai, Aug 5 2026, mini-swe-agent bash-only harness): SWE-bench Verified 88.8%, for the DeepSeek-V4-Flash-0731 build, matching the… full note → |
| 11 | Claude Opus 4.8 ↗Anthropic | 88.6%independent (vals.ai) · vendor ~86% | 69.2% | $5/1M | The hardest agentic refactors and long, autonomous multi-file tasks where every point of accuracy saves a human review cycle.Independent (vals.ai, Jul 14 2026, mini-swe-agent bash-only harness): SWE-bench Verified 88.6% ±1.42. Corrected Jul 17, 2026: we previously printed… full note → |
| 12 | Grok 4.5 ↗SpaceXAI (xAI) | 86.6%independent (vals.ai) | 64.7% | $2/1M | The best value at the top of the board: it solves a SWE-bench Verified task for about $2.31 of input, less than half what the two models above it cost, and it is the fastest of the leaders.Independent (vals.ai, Jul 14 2026, mini-swe-agent bash-only harness): SWE-bench Verified 86.6% ±1.52. Verified Jul 17, 2026 — it launched Jul 8 with… full note → |
| 13 | Muse Spark 1.2 ↗Meta | 86.6%independent (vals.ai) | — | $1.25/1M | Bounded, well-specified changes at low cost, especially if you accept the contributor tier and let Meta train on your traffic.Independent (vals.ai, Aug 6 2026, mini-swe-agent bash-only harness): SWE-bench Verified 86.6%. Added Aug 7, 2026, the day after Meta launched it… full note → |
| 14 | Qwen3.8-27B open ↗Alibaba | 86.0%independent (vals.ai) | 61.7% | — | The best independently measured open-weight model under a true OSI licence, and small enough to serve on a single accelerator.Independent (vals.ai, 2026-08-20 sweep, mini-swe-agent bash-only harness): SWE-bench Verified 86.0% ±1.55, 14th of 86 systems. Left the verifying… full note → |
| 15 | Qwen3.8-Max ↗Alibaba | 85.6%independent (vals.ai) | — | $2.00/1M | Alibaba's 2.4T-parameter flagship, pitched at long-horizon autonomous coding. Independently scored, and slow: the highest latency of any ranked model here.Independent (vals.ai, benchmark updated 2026-08-08, mini-swe-agent bash-only harness): SWE-bench Verified 85.6% ± 1.57. Announced Aug 3, 2026 as "a… full note → |
| 16 | Claude Sonnet 5 ↗Anthropic | 85.2%vendor-reported · vals.ai 79.6% | 63.2% | $2/1M | The best closed-model value — near-Opus scores at ~2.5× less, and the default daily driver for most developers.Vendor-reported (Anthropic), on Anthropic's own scaffold. Independent comparison: vals.ai's bash-only harness measures Sonnet 5 at 79.6% ±1.80, 5.6… full note → |
| 17 | GLM 5.2 open ↗Z.ai (Zhipu AI) | 82.8%independent (vals.ai) | — | $1.40/1M | The best open-weight coder on this board that you can actually download today: an independently measured 82.8%, MIT-licensed, and roughly a third the input price of the closed models above it.Independent (vals.ai, Jul 22 2026, mini-swe-agent bash-only harness): SWE-bench Verified 82.8% ±1.69, 9th of 75 systems. Added Jul 27, 2026 after a… full note → |
| 18 | GPT-5.5 ↗OpenAI | 82.6%independent (vals.ai) | 58.6% | $5/1M | OpenAI's strongest agentic coder, with the deepest tooling and ecosystem breadth of the closed labs.Verified score from vals.ai independent eval; Pro is OpenAI-reported (rivals flag possible memorization on Pro). Price: OpenAI list $5/$30 per 1M… full note → |
| 19 | Muse Spark 1.1 ↗Meta | 82.0%independent (vals.ai) | — | $1.25/1M | Meta's first paid model, and a genuine value pick: a top-10 verified score for $1.52 per solved task, cheaper per result than every model ranked above it.Independent (vals.ai, Jul 14 2026, mini-swe-agent bash-only harness): SWE-bench Verified 82.0% ±1.72. Verified Jul 17, 2026; Meta published no… full note → |
| 20 | Gemini 3.7 Flash ↗Google DeepMind | 80.8%independent (vals.ai) | — | $0.75/1M | Google's volume tier at half the price of 3.6 Flash, and the first 3.7-series model with an independent coding score. A real gain over 3.6 Flash, unlike the last Flash bump.Independent (vals.ai, mini-swe-agent bash-only harness): SWE-bench Verified 80.80% ±1.76. Released Aug 13, 2026, three weeks after Gemini 3.6 Flash… full note → |
| 21 | DeepSeek V4 Pro open ↗DeepSeek | 80.6%vendor-reported · vals.ai 96.4% | 55.4% | $0.435/1M | The cheapest frontier-class coder at ~11× less than Opus, ranked on DeepSeek's own 80.6% although vals.ai measures it at 96.4%. Best pick when cost or self-hosting rules.Vendor-reported: DeepSeek's own model card, Pro-Max mode (SWE-bench Verified 80.6%, SWE-bench Pro 55.4%). Updated Aug 12, 2026: DeepSeek shipped the… full note → |
| 22 | Gemini 3.1 Pro ↗Google DeepMind | 80.6%vendor-reported | 54.2% | $2/1M | Google's strongest coding model today, with deep Workspace/Cloud integration. (A 3.5 Pro is expected but not shipped.)Vendor-reported (DeepMind) pass rate. No independent eval of this exact model; vals.ai has run Gemini 3.1 Pro Preview (02/26) at 78.8%, a preview… full note → |
| 23 | MiniMax M3 open ↗MiniMax | 80.5%vendor-reported · vals.ai 75.0% | 59.0% | $0.60/1M | Open weights with 1M context, multimodal input and computer use — beats GPT-5.5 on SWE-bench Pro at 5–10% of the cost.Vendor-reported at launch (Jun 1, 2026). Independent comparison: vals.ai's bash-only harness measures MiniMax-M3 at 75.0% ±1.94, 5.5 points lower, so… full note → |
| 24 | Qwen3.7 Max ↗Alibaba | 80.4%vendor-reported · vals.ai 68.8% | 60.6% | $1.25/1M | The best non-Claude score on the hardest benchmark — 60.6% SWE-bench Pro — built for long-horizon coding agents.Vendor-reported (May 20, 2026), and the widest vendor-versus-independent gap on this board: vals.ai's bash-only harness measures Qwen 3.7 Max at… full note → |
| 25 | Kimi K2.6 open ↗Moonshot AI | 80.2%vendor-reported · vals.ai 76.2% | 58.6% | $0.95/1M | A top-three open coder whose 58.6% SWE-bench Pro beats several closed flagships.Vendor-reported (10-run average on Moonshot's own SWE-agent harness). Independent comparison: vals.ai's bash-only harness measures Kimi K2.6 at 76.2%… full note → |
| 26 | Gemini 3.8 Flash ↗Google DeepMind | 80.0%independent (vals.ai) | — | $0.75/1M | Google's third Flash release in six weeks, officially launched Sept 2, 2026. Now has an independent score, landing in a tight cluster with Kimi K2.6 and Gemini 3.6 Flash rather than clearly ahead of either.Independent (vals.ai, observed 2026-09-03, mini-swe-agent bash-only harness): SWE-bench Verified 80.00% ±1.79. Leaves the verifying queue it entered… full note → |
| 27 | Gemini 3.6 Flash ↗Google DeepMind | 79.6%independent (vals.ai) | — | $1.50/1M | Google's new general workhorse: it now outscores 3.5 Flash on an independent coding eval while emitting 17% fewer output tokens.Independent (vals.ai, mini-swe-agent bash-only harness): SWE-bench Verified 79.60% ±1.80. Verified Jul 23, 2026 — vals.ai had not run it at our Jul… full note → |
| 28 | Composer 2.5 ↗Cursor | 79.6%independent (vals.ai) | — | $0.50/1M | Cheap, fast in-editor coding if you already pay for Cursor. There is no way to use it anywhere else, which is the whole catch.Independent (vals.ai, Jul 22 2026, mini-swe-agent bash-only harness): SWE-bench Verified 79.6% ±1.80, 13th of 75 systems. Added Jul 27, 2026 from a… full note → |
| 29 | Gemini 3.5 Flash ↗Google DeepMind | 78.8%independent (vals.ai) | — | $1.50/1M | Frontier-ish coding at Flash speed and price, with computer use built in as a native tool.vals.ai independent eval. See our decode of its native computer-use tool. Price: Google list $1.50/$9 per 1M (cached input $0.15). |
| 30 | Gemini 3.5 Flash-Lite ↗Google DeepMind | 75.0%independent (vals.ai) | 54.2% | $0.30/1M | The cheapest ranked model on the board: 75.0% independently measured at $0.30 per 1M input, roughly a fifth the price of the Flash tier above it.Independent (vals.ai, mini-swe-agent bash-only harness): SWE-bench Verified 75.00% ±1.94. Verified Jul 23, 2026 — vals.ai had not run it at our Jul… full note → |
Newest models, not yet independently ranked
12 models have no SWE-bench Verified score from the maker or an independent evaluator, so they are not ranked and not estimated. Figures shown are the makers' own, on other benchmarks, and are not comparable to the ranked scores. Newest first.
| Model | Announced | Terminal-Bench | SWE-bench Pro | Input | Status |
|---|---|---|---|---|---|
| GPT-6.1 Sol ↗OpenAI | - | - | $2/1M | Near-flagship coding and agentic work at a fifth of GPT-6 Astra's per-token cost, aimed at high-volume agent loops rather than one-off chat.Vendor-reported (OpenAI, Sept 29 2026): matches GPT-6 Astra on DeepSWE v1.1 (a real-codebase coding benchmark) at roughly a fifth of the cost, and… | |
| Claude Sonnet 5.5 ↗Anthropic | 70.6 (TB4)vendor-reported | - | $2/1M | Agentic coding at Sonnet-tier pricing: scores higher than Opus 5.5 on Anthropic's own Terminal-Bench 4.0 (70.6% vs 66.4%) while billing at half Opus 5.5's per-token rate, but like Opus 5.5 it ships with no SWE-bench Verified score to rank on.Vendor-reported (Anthropic, Sept 28 2026): Terminal-Bench 4.0 70.6%, FrontierCode 1.1 46.2% (max effort), CursorBench 4.0 55.5%, GDPval-AA v2.1 1844… | |
| Claude Opus 5.5 ↗Anthropic | 66.4 (TB4)vendor-reported | - | $4/1M | Agentic and computer-use work rather than issue-patching: Anthropic leads its own benchmark table with Terminal-Bench 4.0 and OSWorld rather than SWE-bench, and it undercuts Opus 5 on list price while running faster.Vendor-reported (Anthropic, Sept 22 2026): Terminal-Bench 4.0 66.4%, FrontierCode v1.1 54.4%, CursorBench 4.0 57.8%, GDPval-AA v2.1 1846 Elo… | |
| Grok 4.7 ↗SpaceXAI (xAI) | - | - | $2/1M | Launched Sep 21, 2026, targeting longer multi-hour agentic coding sessions with self-verification. No SWE-bench Verified score yet, independent or vendor.Released Sep 21, 2026, about nine days after Musk teased a Sep 11-12 target. SpaceXAI describes a new, larger base model with extended RL training… | |
| Step 5 Preview ↗StepFun | 33.3 (TB4)vendor-reported | - | $1.00/1M | Cheap long-context agentic work with a finance tilt: 600B MoE with 27B active, 1M context, $1 per 1M input, and vendor benchmarks that lead on FrontierFinance and DRACO while trailing the US frontier on terminal and agent tasks.Vendor-reported (StepFun, Sept 20 2026): Terminal-Bench v4 33.3% (vs Claude Opus 5 52.3%, GPT-6 Astra 57.9%), DeepSWE v1.1 67.7%, ProgramBench 80.5%… | |
| Cognition SWE-2 ↗Cognition | - | - | - | Launched Sept 10, 2026, post-trained from Kimi K3 (2.8T params) with a cost-penalized RL reward across three effort tiers (medium/high/max). Nothing to rank on yet: independent.Released Sept 10, 2026. Cognition's own numbers, all on benchmarks Cognition built or controls (FrontierCode 1.1 Main, DeepSWE 1.1, Terminal-Bench… | |
| Muse Glimmer open ↗Meta | - | - | Free/1M | A 30B dense agentic model small enough to run always-on inside a 24GB consumer GPU, distilled from Muse Spark. Nothing to rank on yet.Released Aug 10, 2026 with weights on Hugging Face under a plain Apache 2.0 license, Meta's first genuinely permissive open-weight release in the… | |
| Laguna S 2.1 open ↗Poolside | 70.2% (TB2.1)vendor-reported | 59.4%vendor-reported | $0.10/1M | Frontier-adjacent agentic coding at small-model cost: a 118B mixture-of-experts with only ~8B active per token, open weights, and INT4/GGUF builds that run on local hardware.Vendor-reported (Poolside, Jul 21 2026): Terminal-Bench 2.1 70.2% with thinking enabled (60.4% without), SWE-bench Pro 59.4%, SWE-bench Multilingual… | |
| LongCat-2.0 open ↗Meituan | 70.8%vendor-reported | 59.5%vendor-reported | $0.30/1M | A 1.6T Mixture-of-Experts agentic coder trained entirely on domestic Chinese chips; led OpenRouter usage in stealth as "Owl Alpha".Vendor-reported (Meituan, Jun 30 2026): SWE-bench Pro 59.5%, Terminal-Bench 70.8%. No SWE-bench Verified score or independent eval published yet, so… | |
| MiMo-V2.6-Pro open ↗Xiaomi | - | - | - | $0.43/1M | Teams that need frontier-adjacent capability under a genuinely permissive licence: a 1T-parameter MoE with 42B active, 1M context, omnimodal input, MIT weights, at roughly a tenth of closed frontier list pricing.No SWE-bench Verified score exists for this model from Xiaomi or any independent evaluator, so it enters the verifying queue unranked; we do not… |
| Cohere North Mini Code open ↗Cohere | - | - | - | Free/1M | Private, self-hosted agentic coding for enterprises that cannot send source to a cloud API; runs on a single H100.Open-weight ~30B mixture-of-experts coder, free to use, runs on a single H100 with 256K context. Cohere published no SWE-bench Verified/Pro or… |
| KAT-Coder-Pro V2.5 ↗Kwaipilot (Kuaishou) | - | - | 65.2%vendor-reported | $0.74/1M | Long-horizon agentic coding on a budget: a 72B-active MoE tuned for tool use, with a cheaper Air tier at $0.15/$0.60 on the same 256K context.Vendor-reported (KAT-Coder-V2.5 technical report, arXiv 2607.05471): SWE-bench Pro 65.2%, second only to Opus 4.8 at 69.2%, plus a best-in-test… |
Recent changes
- GPT-6.1 Sol announced, listed unranked No SWE-bench Verified score from its maker or an independent evaluator yet.
- Ember-1 enters the ranking Ranked #9 at 92.2% SWE-bench Verified.
- Claude Sonnet 5.5 announced, listed unranked No SWE-bench Verified score from its maker or an independent evaluator yet.
- Claude Opus 5.5 announced, listed unranked No SWE-bench Verified score from its maker or an independent evaluator yet.
- Grok 4.7 announced, listed unranked No SWE-bench Verified score from its maker or an independent evaluator yet.
- Step 5 Preview announced, listed unranked No SWE-bench Verified score from its maker or an independent evaluator yet.
- Cognition SWE-2 announced, listed unranked No SWE-bench Verified score from its maker or an independent evaluator yet.
- vals.ai archives SWE-bench Verified The independent board this ranking leans on stops scoring new models, citing score saturation. Its 88 systems are frozen, so models released after this may never get an independent score.
- Gemini 3.8 Flash enters the ranking Ranked #25 at 80.0% SWE-bench Verified.
- GPT-5.6 Terra score revised SWE-bench Verified 75.2% to 95.4%, now #4.
how we rank: swe-bench verified, swe-bench pro and terminal-bench
We rank by SWE-bench Verified (500 real, human-validated GitHub issues resolved end-to-end), tiebroken by the harder SWE-bench Pro. A score is only printed once confirmed against an independent evaluation or the maker's primary source — and every row states which kind it is. Where both exist, we print both: one as the ranked score, the other beside it in the table and in that row's note. We would rather show you the gap than ask you to trust our pick.
Our independent reference is vals.ai, which runs every model itself through the same minimal bash-only harness (mini-swe-agent), so the models are compared on equal footing. That matters more than it sounds: SWE-bench scores a model and its scaffolding together, and vendors report using their own tuned scaffolds.
Against vals.ai's neutral harness, the vendor claims on this board usually run optimistic, by 4.0 to 11.6 points, though two run below the independent figure; the vendor-versus-independent chart on this page plots every pair. So rows marked vendor-reported are best-case numbers and are not strictly comparable to the independent ones — where we know the independent figure, we print it beside the score and in the row's note.
That is a deliberate choice and you should know we made it: on five rows (Claude Sonnet 5, MiniMax M3, Qwen3.7 Max, Kimi K2.6, DeepSeek V4 Pro) an independent score for that exact model exists and disagrees with the maker's, and we still rank on the maker's published figure because it is the number that model is sold and quoted on. On four of those the independent number is lower.
DeepSeek V4 Pro runs the other way: vals.ai measures the 0813 GA build 15.8 points ABOVE DeepSeek's own claim, which the disclose-don't-rebase rule was never written for, so that row is flagged in its note for a human decision on whether to rebase it. Until that decision is made it stays ranked on DeepSeek's 80.6%, which is the position the table shows, with vals.ai's 96.4% printed beside it.
We disclose the independent one beside it instead of quietly restating the board on a single evaluator's harness choice. The honest consequence: positions that straddle the two regimes are approximate. Qwen3.7 Max is the sharpest case: it is ranked on Alibaba's 80.4%, and on the neutral harness its 68.8% would put it below every other ranked model.
Note that llm-stats, which we previously miscredited as an independent tracker, labels its own SWE-bench Verified table "Verified: 0 / Self-reported: 104"; it aggregates vendor claims. Models with no SWE-bench Verified score from either the maker or an independent evaluator are not ranked and not estimated.
They are listed separately as the newest models not yet independently ranked, with any figures their makers published on other benchmarks (Terminal-Bench, SWE-bench Pro) labelled vendor-reported, because those numbers are not comparable to the ranked ones. vals.ai archived its SWE-bench Verified board on 5 September 2026, so some of these models may never receive an independent score, and we will not move them into the ranking on a different benchmark's number.
A model whose maker published a SWE-bench Verified score that no independent evaluator has re-run (Ember-1 and Gemini 3.1 Pro today) is ranked on that figure and labelled vendor-reported in the table. Prices are per 1M input tokens on the standard API tier and can change — always confirm current pricing with the provider.
our picks: the best ai models for coding right now
Best overallClaude Opus 5
The top independently measured score on SWE-bench Verified at 97.0%, and it costs half what Fable 5 does. Treat the lead over GPT-5.6 Sol and Fable 5 as a tie, but there is no reason to pay more for the same tier.
Best value (closed)Claude Sonnet 5
85.2% SWE-bench Verified at $2/1M — near-frontier coding at a rounding-error price. The default for most work.
Best open weightsQwen3.8-27B
86.0% Verified measured independently by vals.ai, Apache 2.0 weights on Hugging Face, and a dense 27B that serves on a single accelerator. It overtook GLM 5.2 (82.8%, MIT) on the 2026-08-20 sweep. Kimi K3 scores higher at 93.4% but ships 4-bit only under a bespoke non-OSI licence, so it is open for inference rather than open outright. GLM-5.3 scores far higher still at 95.4%, but Z.ai has not released its weights yet.
Best for autonomous agentsGPT-5.6 Sol
Best on the tasks that actually run long: 98% of the 1-to-4-hour SWE-bench tier, ahead of both Claude Opus 5 (90%) and Fable 5 (93%), which is what matters for unattended repo-wide work.
Best free optionClaude Sonnet 5
It's the default model for free claude.ai users — frontier-class coding at no cost for everyday tasks.
Hardest-tasks dark horseQwen3.7 Max
Its 60.6% on SWE-bench Pro is the best non-Claude score on the benchmark that's hardest to game.
compare head-to-head
how the field got here
- 2021GitHub Copilot preview Autocomplete-in-the-editor goes mainstream.
- 2023ChatGPT + GPT-4, then Cursor Chat-based coding and the first AI-native editor arrive.
- Aug 2024SWE-bench Verified launches A human-validated benchmark of real GitHub issues sets an honest bar.
- Oct 2024Claude 3.5 Sonnet hits ~49% Agents begin resolving real issues, not just snippets.
- 2025Terminal agents Claude Code and Codex CLI move AI out of the editor into the whole repo.
- Apr–Jun 2026Open weights close the gap DeepSeek V4, Kimi K2.6 and peers cluster at ~80% Verified — for pennies.
- 2026Verified saturates in the mid-80s SWE-bench Pro and Terminal-Bench become the real differentiators.
- Jul 2026Claude Fable 5 returns Restored after a 20-day export-control suspension; retakes SWE-bench Pro at 80.3%.
- Jun 30 2026Meituan open-sources LongCat-2.0 A 1.6T MoE coder trained entirely on domestic Chinese chips; vendor-reported 59.5% SWE-bench Pro, awaiting independent eval.
- Jul 25 2026Claude Opus 5 takes #1 vals.ai measures 97.00% ±0.76 on the neutral harness, one day after launch. A statistical tie with GPT-5.6 Sol and Claude Fable 5.
- Sep 5 2026vals.ai archives SWE-bench Verified The independent board this ranking leans on stops scoring new models, citing score saturation. Its 88 systems are frozen, so models released after this may never get an independent score.
Vendor claims vs independent re-runs
Vendor SWE-bench Verified claims ran above independent re-runs on 4 of 6 models, by up to 11.6 points; on 2 they ran below, by up to 15.8 points.
Primary sources
- BenchmarkSWE-bench — the real-GitHub-issue benchmark & leaderboard
- Benchmarkvals.ai — SWE-bench Verified — independent third-party evaluations
- BenchmarkSWE-bench Pro (Scale) — the harder long-horizon leaderboard
- PaperSWE-bench (arXiv) — how the benchmark is constructed
- PressGENZ TECH — Claude Fable 5 returns — restoration & SWE-bench Pro lead
- MakerAnthropic — news — Claude Opus 4.8 / Sonnet 5 model cards & pricing
- MakerOpenAI — GPT-5.5 — GPT-5.5 release post
- MakerMoonshot AI — Kimi K2.6 — open weights & reported scores
- MakerGoogle DeepMind — Gemini — Gemini model pages
- PressVentureBeat — MiniMax-M3 debut — M3 launch scores & pricing
- BenchmarkW&B ml-news — Qwen3.7-Max scores — Qwen3.7-Max benchmark table
- PressGENZ TECH — Meituan LongCat-2.0 — 1.6T MoE coder on Chinese chips (vendor-reported)
cite & embed
Free to cite and embed with attribution. Raw data: JSON · CSV. Embedding this live table adds it to your site and credits GENZ TECH.
Cite
GENZ TECH. (2026). AI Coding Leaderboard. https://genztech.blog/ai-coding-leaderboard/Embed
<iframe src="https://genztech.blog/ai-coding-leaderboard/embed/" width="100%" height="470" loading="lazy" title="AI Coding Leaderboard by GENZ TECH" style="border:1px solid #26282b;border-radius:10px;max-width:560px"></iframe>