the ranking
| # | Model | SWE-bench Verified | Best for |
|---|---|---|---|
| 1 | GPT-5.6 SolOpenAI | 96.2% | The highest independently measured coding score of any model, and it holds up on the long tasks: 98% on the 1-to-4-hour tier where most models fall apart. |
| 2 | Claude Fable 5Anthropic | 95.0% | 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. |
| 3 | Claude Opus 4.8Anthropic | 88.6% | The hardest agentic refactors and long, autonomous multi-file tasks where every point of accuracy saves a human review cycle. |
| 4 | Grok 4.5SpaceXAI (xAI) | 86.6% | 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. |
| 5 | Claude Sonnet 5Anthropic | 85.2% | The best closed-model value — near-Opus scores at ~2.5× less, and the default daily driver for most developers. |
| 6 | GPT-5.5OpenAI | 82.6% | OpenAI's strongest agentic coder, with the deepest tooling and ecosystem breadth of the closed labs. |
the top picks, decoded
1. GPT-5.6 Sol — 96.2% SWE-bench Verified
The highest independently measured coding score of any model, and it holds up on the long tasks: 98% on the 1-to-4-hour tier where most models fall apart. Independent (vals.ai, Jul 14 2026, mini-swe-agent bash-only harness): SWE-bench Verified 96.20% ±0.86 — the top score on the board. Verified Jul 17, 2026; it had been unranked since Jun 26 because OpenAI published no SWE-bench number of its own, and it still has not. Read the #1 with care: the 1.2-point lead over Claude Fable 5 (95.00% ±0.98) is inside the combined margin of error (~0.9 sigma, not significant), so the two are a statistical tie and we rank Sol first only because it scored higher. Where it does separate is task length — 98% on 1-4 hour tasks vs 93% for Fable 5. OpenAI's own Terminal-Bench 2.1 claim is 88.8% (Sol) / 91.9% (Sol Ultra). No SWE-bench Pro score published. Pricing $5/$30 per 1M.
2. Claude Fable 5 — 95.0% SWE-bench Verified
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 evaluated at 96.20% ±0.86 on the same harness — a 1.2-point gap that is inside the combined margin of error (~0.9 sigma), so the two are a statistical tie and we rank Sol first only because it scored higher. SWE-bench Pro 80.3% uses Anthropic's own scaffolding and is contested. Restored Jul 1, 2026 after a 20-day export-control suspension. Pricing $10/$50 per 1M.
3. Claude Opus 4.8 — 88.6% SWE-bench Verified
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 Anthropic's own "~86%" and claimed independent evals tracked it within ~1 point, which was wrong — the independent number is 2.6 points higher, and our methodology is to prefer the independent one. Run through the Claude Code harness instead of the bare bash agent, vals.ai measures 85.8%, a reminder that the harness moves these numbers as much as the model does. SWE-bench Pro 69.2% is Anthropic-reported.
how we rank
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 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 run 2.6 to 11.6 points optimistic. 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 in the row's note. That is a deliberate choice and you should know we made it: on four rows (Claude Sonnet 5, MiniMax M3, Qwen3.7 Max, Kimi K2.6) an independent score for that exact model exists and is lower, and we still rank on the maker's published figure because it is the number that model is sold and quoted on. 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 sits at #11 on Alibaba's 80.4%, and on the neutral harness its 68.8% would put it far down the table. 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 still being checked are marked “verifying” and shown without a number rather than estimated. Prices are per 1M input tokens on the standard API tier and can change — always confirm current pricing with the provider.
Want the raw numbers? The full dataset is public: JSON · CSV.
From our full AI Coding Leaderboard (2026-07-17). We only rank scores confirmed against primary sources.