Ranked by Terminal-Bench 4.0 - 23 models
Best AI models for agentic coding
In board data through September 29, 2026, Claude Opus 5.5 ranks first on the official Terminal-Bench 4.0 leaderboard with 64.8% ± 3.1%, 3 points ahead of Claude Sonnet 5.5 at 61.8% ± 2.9%. That gap is within the ±3.1 margin the board prints on the leading score. Its list price is $4 input / $20 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 Claude Sonnet 5.5 at 61.8% ± 2.9% ($2 input / $10 output per 1M tokens). The highest-placed open-weight model is GLM-5.3, ranked 9th with 41.8% ± 3.2%.
Board data through , checked . How this is ranked
Terminal-Bench 4.0 ranking
| Rank | Model | Provider | Terminal-Bench 4.0 | Source | Input / 1M | Output / 1M | Context | Weights | DeepSWE v1.1thru Sep 22, 2026 |
|---|---|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 5.5Board: Opus 5.5 | Anthropic | 64.8% ± 3.1%Official leaderboard - Rank 1; agent: Claude Code; effort: max; tokens 8.0B; cost $4.7k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $4 | $20 | 1M | Closed | - |
| 2 | Claude Sonnet 5.5Board: Sonnet 5.5 | Anthropic | 61.8% ± 2.9%Official leaderboard - Rank 2; agent: Claude Code; effort: max; tokens 19.4B; cost $7.3k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $2 | $10 | 1M | Closed | - |
| 3 | GPT-6 Astra | OpenAI | 58.2% ± 2.8%Official leaderboard - Rank 3 (tie); agent: Codex; effort: max; tokens 1.5B; cost $3.3k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $10 | $50 | 1.05M | Closed | 74%±3% |
| 3 | GPT-6.1 Sol | OpenAI | 58.2% ± 3.1%Official leaderboard - Rank 3 (tie); agent: Codex; effort: max; tokens 1.5B; cost $0.6k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $2 | $10 | 1.05M | Closed | - |
| 5 | Claude Fable 5.1Board: Fable 5.1 | Anthropic | 57.9% ± 3.8%Official leaderboard - Rank 5; agent: Claude Code; effort: max; tokens 2.7B; cost $6.2k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $10 | $50 | 1M | Closed | - |
| 6 | Claude Opus 5Board: Opus 5 | Anthropic | 53.9% ± 3.2%Official leaderboard - Rank 6; agent: Claude Code; effort: xhigh; tokens 6.9B; cost $6.1k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $5 | $25 | 1M | Closed | 74%±4% |
| 7 | GPT-6 Sol | OpenAI | 49.4% ± 3.2%Official leaderboard - Rank 7; agent: Codex; effort: max; tokens 5.0B; cost $1.5k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $2 | $10 | 1.05M | Closed | - |
| 8 | Claude Fable 5Board: Fable 5 | Anthropic | 44.5% ± 3.8%Official leaderboard - Rank 8; agent: Claude Code; effort: max; tokens 3.8B; cost $7.3k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $10 | $50 | 1M | Closed | 70%±3% |
| 9 | GLM-5.3 | Z.ai | 41.8% ± 3.2%Official leaderboard - Rank 9; agent: Claude Code; effort: max; tokens 8.7B; cost $2.7k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $1.40 | $4.40 | 1M | Open | 69%±3% |
| 10 | Grok 4.7 | xAI | 37.6% ± 3.5%Official leaderboard - Rank 10; agent: Grok Build; effort: xhigh; tokens 5.5B; cost $3.7k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $2 | $6 | 500K | Closed | - |
| 11 | GPT-5.6 Sol | OpenAI | 37.3% ± 3.8%Official leaderboard - Rank 11; agent: Codex; effort: max; tokens 4.4B; cost $2.5k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $4 | $20 | 1.05M | Closed | 73%±3% |
| 12 | GLM-5.3-Flash | Z.ai | 35.8% ± 3.5%Official leaderboard - Rank 12; agent: Claude Code; effort: none; tokens 2.9B; cost $2.9k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $0.15 | $0.50 | 1M | Open | 63%±4% |
| 13 | Qwen3.8-MaxBoard: Qwen3.8-Max-0902 | Alibaba Cloud (Qwen) | 27.0% ± 3.8%Official leaderboard - Rank 13; agent: Claude Code; effort: max; tokens 3.0B; cost $4.6k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $2 | $6 | 1M | Closed | 57%±3% |
| 14 | Claude Opus 4.8Board: Opus 4.8 | Anthropic | 23.6% ± 3.6%Official leaderboard - Rank 14; agent: Claude Code; effort: max; tokens 6.4B; cost $6.5k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $5 | $25 | 1M | Closed | 59%±2% |
| 15 | GPT-5.6 Terra | OpenAI | 21.5% ± 3.3%Official leaderboard - Rank 15; agent: Codex; effort: max; tokens 5.7B; cost $1.7k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $2 | $12 | 1.05M | Closed | - |
| 16 | Grok 4.6 | xAI | 20.3% ± 3.1%Official leaderboard - Rank 16; agent: Grok Build; effort: high; tokens 4.0B; cost $3.6k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $2 | $6 | 500K | Closed | 67%±2% |
| 17 | Gemini 3.8 Flash | 19.1% ± 3.4%Official leaderboard - Rank 17; agent: mini-SWE-agent; effort: high; tokens 17.2B; cost $1.8k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $0.75 | $3.75 | 1.05M | Closed | 74%±1% | |
| 18 | GPT-5.6 Luna | OpenAI | 17.3% ± 2.8%Official leaderboard - Rank 18; agent: Codex; effort: max; tokens 11.6B; cost $0.3k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $0.20 | $1.20 | 1.05M | Closed | 67%±4% |
| 19 | GPT-6 Luna | OpenAI | 16.4% ± 2.7%Official leaderboard - Rank 19; agent: Codex; effort: max; tokens 3.9B; cost $0.1k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $0.10 | $0.50 | 1.05M | Closed | - |
| 20 | Muse Spark 1.3 | - | 14.5% ± 2.9%Official leaderboard - Rank 20; agent: Muse Code; effort: xhigh; tokens 9.0B; cost not shown; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | - | - | - | - | - |
| 21 | Grok 4.5 | xAI | 12.4% ± 2.6%Official leaderboard - Rank 21 (tie); agent: Grok Build; effort: high; tokens 3.4B; cost $2.1k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $2 | $6 | 500K | Closed | - |
| 21 | Claude Sonnet 5Board: Sonnet 5 | Anthropic | 12.4% ± 3.1%Official leaderboard - Rank 21 (tie); agent: Claude Code; effort: max; tokens 21.6B; cost $9.6k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | $2 | $10 | 1M | Closed | 54%±4% |
| 23 | Gemini 3.7 Flash | - | 11.2% ± 2.4%Official leaderboard - Rank 23; agent: mini-SWE-agent; effort: high; tokens 11.1B; cost $1.3k; date = model release date shown on leaderboard; read 2026-10-17 | tbench.ai | - | - | - | - | 65%±3% |
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 agentic coding?
Claude Opus 5.5 ranks first by Terminal-Bench 4.0, board data through September 29, 2026, with 64.8% ± 3.1% (official leaderboard). 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 agentic coding near the top of Terminal-Bench 4.0?
Taking every ranked model within 5 points of the top score that has a price in our registry, Claude Sonnet 5.5 has the lowest output price: $2 input / $10 output per 1M tokens, with 61.8% ± 2.9% by Terminal-Bench 4.0, board data through September 29, 2026.
What is the best open-weight model for agentic coding?
GLM-5.3 is the highest-placed open-weight model by Terminal-Bench 4.0, board data through September 29, 2026, ranked 9th with 41.8% ± 3.2%.
How is this agentic coding ranking built?
Rows follow the Terminal-Bench 4.0 board's own order, checked on October 17, 2026; its newest entry is dated September 29, 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.