Ranked by DeepSWE v1.1 - 21 models
Best AI models for coding
In board data through September 22, 2026, GPT-6 Astra (74%±3%), Gemini 3.8 Flash (74%±1%) and Claude Opus 5 (74%±4%) are tied for first on the official DeepSWE v1.1 leaderboard, 1 point ahead of GPT-5.6 Sol at 73%±3%. That gap is within the ±3 margin the board prints on the leading score. GPT-6 Astra lists at $10 input / $50 output per 1M tokens; Gemini 3.8 Flash lists at $0.75 input / $3.75 output per 1M tokens; Claude Opus 5 lists at $5 input / $25 output per 1M tokens. A fixed rule, the lowest output price per 1M tokens among models with a registry price that score within 5 points of the top score, picks Gemini 3.8 Flash at 74%±1% ($0.75 input / $3.75 output per 1M tokens). The highest-placed open-weight model is GLM-5.3, ranked 6th with 69%±3%.
Board data through , checked . How this is ranked
DeepSWE v1.1 ranking
| Rank | Model | Provider | DeepSWE v1.1 | Source | Input / 1M | Output / 1M | Context | Weights | SWE-bench Verifiedthru Feb 17, 2026 | SWE-Bench Proundated | LiveCodeBenchthru Aug 1, 2025 | Aider Polyglotthru Oct 3, 2025 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | GPT-6 AstraBoard: gpt-6-astra | OpenAI | 74%±3%Official leaderboard - mini-swe-agent, xhigh effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $4.43/task | deepswe.datacurve.ai | $10 | $50 | 1.05M | Closed | - | - | - | - |
| 1 | Gemini 3.8 FlashBoard: gemini-3.8-flash | 74%±1%Official leaderboard - mini-swe-agent, high effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $2.36/task | deepswe.datacurve.ai | $0.75 | $3.75 | 1.05M | Closed | - | - | - | - | |
| 1 | Claude Opus 5Board: claude-opus-5 | Anthropic | 74%±4%Official leaderboard - mini-swe-agent, max effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $11.84/task | deepswe.datacurve.ai | $5 | $25 | 1M | Closed | - | - | - | - |
| 4 | GPT-5.6 SolBoard: gpt-5.6-sol | OpenAI | 73%±3%Official leaderboard - mini-swe-agent, max effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $6.46/task | deepswe.datacurve.ai | $4 | $20 | 1.05M | Closed | - | - | - | - |
| 5 | Claude Fable 5Board: claude-fable-5 | Anthropic | 70%±3%Official leaderboard - mini-swe-agent, xhigh effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $13.41/task | deepswe.datacurve.ai | $10 | $50 | 1M | Closed | - | - | - | - |
| 6 | GLM-5.3Board: glm-5.3 | Z.ai | 69%±3%Official leaderboard - mini-swe-agent, max effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $3.99/task | deepswe.datacurve.ai | $1.40 | $4.40 | 1M | Open | - | - | - | - |
| 6 | Kimi K3Board: kimi-k3 | Moonshot AI | 69%±5%Official leaderboard - mini-swe-agent, max effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $4.65/task | deepswe.datacurve.ai | $3 | $15 | 1.05M | Open | - | - | - | - |
| 8 | Grok 4.6Board: grok-4.6 | xAI | 67%±2%Official leaderboard - mini-swe-agent, medium effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $3.45/task | deepswe.datacurve.ai | $2 | $6 | 500K | Closed | - | - | - | - |
| 8 | GPT-5.6 LunaBoard: gpt-5.6-luna | OpenAI | 67%±4%Official leaderboard - mini-swe-agent, max effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $0.61/task | deepswe.datacurve.ai | $0.20 | $1.20 | 1.05M | Closed | - | - | - | - |
| 8 | GPT-5.5Board: gpt-5.5 | OpenAI | 67%±6%Official leaderboard - mini-swe-agent, xhigh effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $7.23/task | deepswe.datacurve.ai | $5 | $30 | 1.05M | Closed | - | - | - | - |
| 11 | gemini-3.7-flash | - | 65%±3%Official leaderboard - mini-swe-agent, medium effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $2.03/task | deepswe.datacurve.ai | - | - | - | - | - | - | - | - |
| 12 | GLM-5.3-FlashBoard: glm-5.3-flash | Z.ai | 63%±4%Official leaderboard - mini-swe-agent, max effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $0.24/task | deepswe.datacurve.ai | $0.15 | $0.50 | 1M | Open | - | - | - | - |
| 12 | DeepSeek V4 ProBoard: deepseek-v4-pro | DeepSeek | 63%±6%Official leaderboard - mini-swe-agent, max effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $1.67/task | deepswe.datacurve.ai | $1.32 | $3.96 | 1M | Open | - | - | - | - |
| 14 | Claude Opus 4.8Board: claude-opus-4.8 | Anthropic | 59%±2%Official leaderboard - mini-swe-agent, max effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $13.22/task | deepswe.datacurve.ai | $5 | $25 | 1M | Closed | - | - | - | - |
| 15 | Qwen3.8-MaxBoard: qwen3.8-max | Alibaba Cloud (Qwen) | 57%±3%Official leaderboard - mini-swe-agent, xhigh effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $3.73/task | deepswe.datacurve.ai | $2 | $6 | 1M | Closed | - | - | - | - |
| 16 | muse-spark-1.2 | - | 55%±2%Official leaderboard - mini-swe-agent, xhigh effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $3.70/task | deepswe.datacurve.ai | - | - | - | - | - | - | - | - |
| 17 | Claude Sonnet 5Board: claude-sonnet-5 | Anthropic | 54%±4%Official leaderboard - mini-swe-agent, max effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $26.40/task | deepswe.datacurve.ai | $2 | $10 | 1M | Closed | - | - | - | - |
| 18 | deepseek-v4-flash | - | 53%±4%Official leaderboard - mini-swe-agent, max effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $0.46/task | deepswe.datacurve.ai | - | - | - | - | - | - | - | - |
| 19 | Gemini 3.6 FlashBoard: gemini-3.6-flash | 47%±4%Official leaderboard - mini-swe-agent, high effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $2.21/task | deepswe.datacurve.ai | $0.75 | $3.75 | 1.05M | Closed | - | - | - | - | |
| 20 | GLM-5.2Board: glm-5.2 | Z.ai | 44%±2%Official leaderboard - mini-swe-agent, max effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $3.92/task | deepswe.datacurve.ai | $1.40 | $4.40 | 1M | Open | - | - | - | - |
| 21 | gemini-3.5-flash | - | 36%±4%Official leaderboard - mini-swe-agent, high effort, Pass@1, v1.1 (113 tasks), best effort level shown; avg cost $3.45/task | deepswe.datacurve.ai | - | - | - | - | - | - | - | - |
Prices in USD per 1M tokens at the standard rate, from each provider's pricing page. A dash means the board name does not match a model in our registry or the model is not on that board. Each related-board score links to its source.
Related
API capabilities
Providers of the top 3
Curated picks
FAQ
What is the best AI model for coding?
GPT-6 Astra (74%±3%), Gemini 3.8 Flash (74%±1%) and Claude Opus 5 (74%±4%) are tied for first by DeepSWE v1.1, board data through September 22, 2026. One benchmark is not a verdict: settings differ between rows, so test on your own workload before you switch.
What is the cheapest AI model for coding near the top of DeepSWE v1.1?
Taking every ranked model within 5 points of the top score that has a price in our registry, Gemini 3.8 Flash has the lowest output price: $0.75 input / $3.75 output per 1M tokens, with 74%±1% by DeepSWE v1.1, board data through September 22, 2026.
What is the best open-weight model for coding?
GLM-5.3 is the highest-placed open-weight model by DeepSWE v1.1, board data through September 22, 2026, ranked 6th with 69%±3%.
How is this coding ranking built?
Rows follow the DeepSWE v1.1 board's own order, checked on October 17, 2026; its newest entry is dated September 22, 2026, and a use case is only published while that date is within 120 days of the check. Each model appears once. Prices, context windows and weights come from each provider's first-party pricing page in our model registry; a model the registry does not list shows a dash.
Newsletter
Get new benchmark results as they land
One email per week. New model releases, the leaderboard moves that matter and what they mean for what you build.