Current LLM coding leaderboard
A live-generated snapshot of BenchLM’s coding-ranked lane, joined conservatively to OpenRouter’s public API catalog for current API pricing. It is a shortlist, not a universal quality or value verdict; evaluate finalists on your own workload.
Snapshot and ranked lane
- Generated/fetched: 2026-10-15T07:00:36Z (UTC); BenchLM payload last updated: 2026-10-14; BenchLM HTTP fetched: Thu, 15 Oct 2026 07:00:36 GMT; OpenRouter HTTP fetched: Thu, 15 Oct 2026 07:00:36 GMT.
- BenchLM snapshot:
2026-10-14-fb2b0cfe198c57ee; methodology:bench-align-v5.8-2026-09-30; license: CC BY-NC 4.0. Attribution: Data from BenchLM.ai. Non-commercial use only under that license; commercial use requires permission from BenchLM. - This table uses the rank order and
categoryScores.codingvalues from the requested coding endpoint. It includes rows with different source evidence levels (including supported, reported, and estimated); this is not a verified-only lane. Scores and ranks are source-provided, not recalculated here. - BenchLM’s aggregate methodology and underlying benchmark set can change between snapshots. Do not read small score/rank changes as capability changes without checking the methodology version and source evidence.
Leaderboard
🔬 = BenchLM ranking source; ↗ = matched OpenRouter model page.
| # | Model (provider) | BenchLM coding | Input $/1M | Output $/1M | Context | 🔬 | ↗ |
|---|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5 (Anthropic) | 84.14 | $2 | $10 | 1M | 🔬 | ↗ |
| 2 | Claude Opus 5.5 (Anthropic) | 82.58 | $4 | $20 | 1M | 🔬 | ↗ |
| 3 | Claude Fable 5.1 (Anthropic) | 76.78 | $10 | $50 | 1M | 🔬 | ↗ |
| 4 | GPT-6 Astra (OpenAI) | 75.43 | $10 | $50 | 1.05M | 🔬 | ↗ |
| 5 | Muse Wonder Full (Meta) | 73.71 | — | — | — | 🔬 | — |
| 6 | Gemini 4 Argon (Google) | 72.21 | $2 | $10 | 1.05M | 🔬 | ↗ |
| 7 | Claude Fable 5 (Anthropic) | 71.24 | $10 | $50 | 1M | 🔬 | ↗ |
| 8 | GPT-6.1 Sol (OpenAI) | 70.89 | $2 | $10 | 1.05M | 🔬 | ↗ |
| 9 | Claude Opus 5 (Anthropic) | 70.66 | $5 | $25 | 1M | 🔬 | ↗ |
| 10 | Muse Wonder (Meta) | 69.58 | $4 | $20 | 1.05M | 🔬 | ↗ |
| 11 | GPT-5.6 Sol (OpenAI) | 67.42 | $4 | $20 | 1.05M | 🔬 | ↗ |
| 12 | GPT-6 Sol (OpenAI) | 66.35 | $2 | $10 | 1.05M | 🔬 | ↗ |
| 13 | MiMo-V2.6-Pro (Xiaomi) | 63.27 | $0.435 | $0.87 | 1.05M | 🔬 | ↗ |
| 14 | GPT-5.5 (OpenAI) | 62.07 | $5 | $30 | 1.05M | 🔬 | ↗ |
| 15 | GPT-5.6 Terra (OpenAI) | 61.94 | $2 | $12 | 1.05M | 🔬 | ↗ |
| 16 | GPT-5.6 Luna (OpenAI) | 61.62 | $0.2 | $1.2 | 1.05M | 🔬 | ↗ |
| 17 | Gemini 3.8 Flash (Google) | 60.85 | $0.75 | $3.75 | 1.05M | 🔬 | ↗ |
| 18 | DeepSeek V4.1 Flash (DeepSeek) | 60.74 | $0.0137 | $0.891 | 1.05M | 🔬 | ↗ |
| 19 | Claude Opus 4.8 (Anthropic) | 60.70 | $5 | $25 | 1M | 🔬 | ↗ |
| 20 | Step 5 Preview (StepFun) | 60.43 | $1 | $2.7 | 1.05M | 🔬 | ↗ |
| 21 | MiMo-V2.6-Flash (Xiaomi) | 60.12 | $0.14 | $0.28 | 1.05M | 🔬 | ↗ |
| 22 | Kimi K3 (Moonshot AI) | 58.98 | $0.45 | $15 | 1.05M | 🔬 | ↗ |
| 23 | Grok 4.6 (xAI) | 58.60 | $2 | $6 | 500K | 🔬 | ↗ |
| 24 | Gemini 3.7 Flash (Google) | 57.59 | $0.75 | $3.75 | 1.05M | 🔬 | ↗ |
| 25 | Claude Sonnet 5 (Anthropic) | 57.52 | $2 | $10 | 1M | 🔬 | ↗ |
| 26 | Ling 3.1 Flash (InclusionAI) | 57.08 | $0.075 | $0.225 | 262K | 🔬 | ↗ |
| 27 | Hy4 preview (Tencent) | 56.71 | $0.834 | $2.501 | 1.05M | 🔬 | ↗ |
| 28 | GLM-5.3 (Z.AI) | 56.44 | $0.079 | $4.8 | 1.05M | 🔬 | ↗ |
| 29 | Grok 4.5 (xAI) | 55.75 | $2 | $6 | 500K | 🔬 | ↗ |
| 30 | Claude Opus 4.7 (Adaptive) (Anthropic) | 55.67 | — | — | — | 🔬 | — |
| 31 | Claude Opus 4.7 (Anthropic) | 55.42 | $5 | $25 | 1M | 🔬 | ↗ |
| 32 | Qwen3.8 Max (Alibaba) | 53.72 | — | — | — | 🔬 | — |
| 33 | GLM-5.2 (Z.AI) | 53.72 | $0.037 | $7 | 1.05M | 🔬 | ↗ |
| 34 | Ornith-1.5-397B (Ornith AI) | 53.67 | — | — | — | 🔬 | — |
| 35 | GPT-5.3 Codex (OpenAI) | 53.59 | $1.75 | $14 | 400K | 🔬 | ↗ |
| 36 | GPT-6 Luna (OpenAI) | 53.25 | $0.1 | $0.5 | 1.05M | 🔬 | ↗ |
| 37 | Claude Haiku 5.5 (Anthropic) | 53.03 | $0.1 | $0.5 | 1M | 🔬 | ↗ |
| 38 | Muse Spark 1.1 (Meta) | 52.67 | $1.25 | $4.25 | 1.05M | 🔬 | ↗ |
| 39 | Muse Spark 1.2 (Meta) | 52.66 | $1.25 | $4.25 | 1.05M | 🔬 | ↗ |
| 40 | Qwen3.8-Flash-Next (Alibaba) | 51.63 | — | — | — | 🔬 | — |
| 41 | Gemini 3.6 Flash (Google) | 51.10 | $0.75 | $3.75 | 1.05M | 🔬 | ↗ |
| 42 | Gemini 3.5 Flash (Google) | 50.14 | $1.5 | $9 | 1.05M | 🔬 | ↗ |
| 43 | Muse Spark (Meta) | 49.90 | — | — | — | 🔬 | — |
| 44 | Claude Opus 4.6 (Adaptive) (Anthropic) | 48.99 | — | — | — | 🔬 | — |
| 45 | GLM-5.3-Flash (Z.AI) | 48.39 | $0.036 | $0.9 | 1.05M | 🔬 | ↗ |
| 46 | Hy3 (Tencent) | 47.96 | $0.132 | $0.528 | 262K | 🔬 | ↗ |
| 47 | GPT-5.4 (OpenAI) | 47.72 | $2.5 | $15 | 1.05M | 🔬 | ↗ |
| 48 | GLM-5.1 (Z.AI) | 47.26 | $0.966 | $3.036 | 205K | 🔬 | ↗ |
| 49 | DeepSeek V4 Pro 0813 (DeepSeek) | 47.24 | $1.32 | $3.96 | 1.05M | 🔬 | ↗ |
| 50 | Qwen3.8-27B (Alibaba) | 46.74 | $0.24 | $2.2 | 1M | 🔬 | ↗ |
| 51 | Claude Opus 4.6 (Anthropic) | 46.13 | $5 | $25 | 1M | 🔬 | ↗ |
| 52 | Mistral Large 4 (Mistral) | 45.53 | $0.68 | $2.09 | 1.05M | 🔬 | ↗ |
| 53 | Claude Sonnet 4.6 (Anthropic) | 44.76 | $3 | $15 | 1M | 🔬 | ↗ |
| 54 | Qwen3.7 Max (Alibaba) | 44.20 | $1.475 | $4.425 | 1M | 🔬 | ↗ |
| 55 | GPT-5.2-Codex (OpenAI) | 43.67 | $1.75 | $14 | 400K | 🔬 | ↗ |
| 56 | Kimi K2.6 (Moonshot AI) | 43.20 | $0.65 | $3.41 | 262K | 🔬 | ↗ |
| 57 | MiMo-V2.5-Pro (Xiaomi) | 43.19 | $0.435 | $0.87 | 1.05M | 🔬 | ↗ |
| 58 | Gemini 3 Pro (Google) | 42.98 | — | — | — | 🔬 | — |
| 59 | Qwen 3.6 Max (preview) (Alibaba) | 42.44 | — | — | — | 🔬 | — |
| 60 | Apodex 1.1 (Apodex) | 42.01 | — | — | — | 🔬 | — |
| 61 | Kimi K2.7 Code (Moonshot AI) | 41.89 | $0.6712 | $3.35 | 262K | 🔬 | ↗ |
| 62 | Quasar 438B (Multiverse Computing) | 41.12 | — | — | — | 🔬 | — |
| 63 | Qwen3.7 Plus (Alibaba) | 40.56 | $0.32 | $1.28 | 1M | 🔬 | ↗ |
| 64 | Claude Opus 4.5 (Anthropic) | 40.37 | $5 | $25 | 200K | 🔬 | ↗ |
| 65 | Qwen3.6 Plus (Alibaba) | 39.97 | $0.325 | $1.95 | 1M | 🔬 | ↗ |
| 66 | Gemini 3.1 Pro (Google) | 39.21 | — | — | — | 🔬 | — |
| 67 | Kimi K2.5 (Moonshot AI) | 38.43 | $0.5 | $2.5 | 262K | 🔬 | ↗ |
| 68 | MiMo-V2-Pro (Xiaomi) | 38.23 | — | — | — | 🔬 | — |
| 69 | GLM-5 (Z.AI) | 37.96 | $1 | $3.2 | 205K | 🔬 | ↗ |
| 70 | Inkling-Small (Thinking Machines Lab) | 37.78 | $0.45 | $1.2 | 524K | 🔬 | ↗ |
| 71 | GLM-4.7 (Z.AI) | 37.38 | $0.6 | $2.2 | 205K | 🔬 | ↗ |
| 72 | GPT-5.2 (OpenAI) | 36.91 | $1.75 | $14 | 400K | 🔬 | ↗ |
| 73 | MiniMax M3 (MiniMax) | 36.49 | $0.3 | $1.2 | 1.05M | 🔬 | ↗ |
| 74 | GPT-5.1 (OpenAI) | 36.12 | $1.25 | $10 | 400K | 🔬 | ↗ |
| 75 | Gemini 3.5 Flash-Lite (Google) | 35.76 | $0.3 | $2.5 | 1.05M | 🔬 | ↗ |
| 76 | GPT-5.4 mini (OpenAI) | 35.19 | $0.75 | $4.5 | 400K | 🔬 | ↗ |
| 77 | Qwen3.6-27B (Alibaba) | 34.79 | $0.32 | $3.2 | 262K | 🔬 | ↗ |
| 78 | MiMo-V2.5 (Xiaomi) | 34.70 | $0.14 | $0.28 | 1.05M | 🔬 | ↗ |
| 79 | MiniMax M2.7 (MiniMax) | 34.47 | $0.3 | $1.2 | 205K | 🔬 | ↗ |
| 80 | MiniMax M2.5 (MiniMax) | 34.07 | $0.27 | $0.95 | 205K | 🔬 | ↗ |
| 81 | Inkling (Thinking Machines Lab) | 33.98 | $1 | $4.05 | 524K | 🔬 | ↗ |
| 82 | GPT-5.1-Codex (OpenAI) | 33.58 | $1.25 | $10 | 400K | 🔬 | ↗ |
| 83 | Qwen3.5-122B-A10B (Alibaba) | 32.68 | $0.26 | $2.08 | 262K | 🔬 | ↗ |
| 84 | GLM-5V-Turbo (Z.AI) | 32.34 | $1.2 | $4 | 203K | 🔬 | ↗ |
| 85 | Gemini 3 Flash (Google) | 32.20 | — | — | — | 🔬 | — |
| 86 | Qwen3.5-27B (Alibaba) | 32.12 | $0.26 | $2.6 | 262K | 🔬 | ↗ |
| 87 | GPT-5 (medium) (OpenAI) | 32.03 | — | — | — | 🔬 | — |
| 88 | Ling 3.0 Flash (InclusionAI) | 31.79 | $0.021 | $0.063 | 262K | 🔬 | ↗ |
| 89 | Gemma 4 31B (Google) | 31.72 | $0.2 | $0.4 | 262K | 🔬 | ↗ |
| 90 | Hy3 Preview (Tencent) | 31.72 | $0.18 | $0.6 | 262K | 🔬 | ↗ |
| 91 | Qwen3.6-35B-A3B (Alibaba) | 31.61 | $0.15 | $1 | 262K | 🔬 | ↗ |
| 92 | Gemma 4 26B A4B (Google) | 30.92 | $0.09 | $0.3 | 262K | 🔬 | ↗ |
| 93 | Muse Glimmer 30B (Meta) | 30.47 | $0.3 | $1.2 | 131K | 🔬 | ↗ |
| 94 | DeepSeek V3.2 (Thinking) (DeepSeek) | 30.25 | — | — | — | 🔬 | — |
| 95 | Step 3.7 Flash (StepFun) | 30.17 | $0.2 | $1.15 | 262K | 🔬 | ↗ |
| 96 | GPT-5.4 nano (OpenAI) | 29.61 | $0.2 | $1.25 | 400K | 🔬 | ↗ |
| 97 | MiMo-V2-Flash (Xiaomi) | 28.50 | — | — | — | 🔬 | — |
| 98 | DeepSeek V3.2 (DeepSeek) | 28.43 | $0.259 | $0.8 | 164K | 🔬 | ↗ |
| 99 | o1 (OpenAI) | 27.61 | $15 | $60 | 200K | 🔬 | ↗ |
| 100 | Gemini 3.1 Flash-Lite (Google) | 27.32 | $0.25 | $1.5 | 1.05M | 🔬 | ↗ |
Prices are OpenRouter’s pricing.prompt and pricing.completion USD per token multiplied by 1,000,000, i.e. standard input/output rates; cache, batch, tool, and long-context tier prices are excluded. Context is OpenRouter’s listed token limit, not a guarantee of effective context. Exact normalized model-name matches with a unique OpenRouter catalog entry only; uncertain or ambiguous matches remain unlinked and unpriced (—). OpenRouter catalog name matching can miss aliases or distinct product variants; verify before purchase.
Interpretation and caveats
- BenchLM ranks its coding aggregate, not a universal measure of coding-agent performance. Its methodology version and the source evidence vary by row; this is not a verified-only lane.
- Scores from SWE-bench and human-preference arenas are intentionally not mixed into this ranking: their tasks, harnesses, populations, and scales differ. A separate result belongs here only when the exact model variant and benchmark setup are independently matched and verifiable.
- Per-1M prices are provider catalog rates as observed at refresh time, not a quote; provider routing, discounts, caches, context tiers, and output/reasoning token accounting can change effective cost.
- Missing data is shown as — rather than inferred. Benchmark inclusion does not imply OpenRouter availability, and an OpenRouter match does not imply benchmark equivalence.
- Model names are linked only when a unique OpenRouter model-page match exists; unmatched names remain plain text. The 🔬 column links to BenchLM’s ranked data for every row; ↗ links to the matched OpenRouter model page.
Changelog
- 2026-10-15: added 4 (Gemini 3.1 Flash-Lite, Ling 3.1 Flash, Muse Wonder, Muse Wonder Full); dropped 4 (GPT-5 mini, Grok 4.3, Laguna M.1, Qwen3.5-35B-A3B); 21 rank moves >3 places; largest: Claude Haiku 5.5 20→37, Muse Spark 1.2 33→39, Qwen3.6-35B-A3B 86→91, Qwen3.5-27B 81→86, Step 3.7 Flash 90→95; 1 coding scores changed >5 points (Claude Haiku 5.5); 13 prices changed >10% or became available/unavailable (DeepSeek V4.1 Flash, GLM-5, GLM-5.2, GLM-5.3-Flash, GPT-5.6 Sol…).
- 2026-10-12: 1 rank moves >3 places; largest: Qwen3.6-35B-A3B 80→86; 13 prices changed >10% or became available/unavailable (DeepSeek V3.2, GLM-5.2, GLM-5.3, Gemma 4 26B A4B, Kimi K2.5…).
- 2026-10-08: added 2 (Claude Haiku 5.5, Mistral Large 4); dropped 2 (GLM-4.6, Grok 4.1 Fast (Reasoning)); 16 rank moves >3 places; largest: MiMo-V2.6-Flash 38→17, GPT-6.1 Sol 22→8, Gemini 3 Flash 75→85, GPT-6 Luna 43→36, Qwen 3.6 Max (preview) 50→57; 3 coding scores changed >5 points (GPT-6 Sol, GPT-6.1 Sol, MiMo-V2.6-Flash); 9 prices changed >10% or became available/unavailable (DeepSeek V4 Pro 0813, GLM-5.1, GLM-5.2, GLM-5.3, Kimi K2.6…).
- 2026-10-05: 13 prices changed >10% or became available/unavailable (DeepSeek V4 Pro 0813, DeepSeek V4.1 Flash, GLM-5.1, GLM-5.2, GLM-5.3…).
- 2026-10-01: added 2 (Gemini 4 Argon, Qwen3.5-35B-A3B); dropped 2 (Gemini 3.1 Flash-Lite, K-Exaone); 16 rank moves >3 places; largest: DeepSeek V4.1 Flash 43→23, GPT-6.1 Sol 8→22, Muse Glimmer 30B 76→87, MiMo-V2-Pro 69→62, GPT-5.1-Codex 85→78; 2 coding scores changed >5 points (DeepSeek V4.1 Flash, GPT-6.1 Sol); 9 prices changed >10% or became available/unavailable (DeepSeek V4 Pro 0813, DeepSeek V4.1 Flash, GLM-5.2, GLM-5.3, Hy4 preview…).
Sources and license
- BenchLM machine-readable coding leaderboard — ranked rows, scores, evidence labels, snapshot and methodology metadata.
- BenchLM dataset and methodology/licensing overview — data coverage, ranking lanes, update date and attribution. Data from BenchLM.ai, licensed CC BY-NC 4.0; commercial use is not permitted without a separate license.
- BenchLM methodology — score construction and verification approach.
- Creative Commons Attribution-NonCommercial 4.0.
- OpenRouter model catalog API — model identifiers, listed context lengths and per-token input/output pricing.
- How to compare coding benchmarks — explanation of benchmark differences and interpretation.