Evaluations and rankings
Compare capability, cost and recent releases. Scores from different tasks and reports are kept separate.
Capability leaders
One highest-scoring configuration per provider.
Latest releases
Models with known release dates; future releases excluded.
- GPT-6.1 TerraFull report (中文)2026-10-26
- kimi-k3.1-proFull report (中文)2026-10-24
- kimi-k3.1-pro-highspeedFull report (中文)2026-10-24
- Voxtral Mini Transcribe 3Full report (中文)2026-10-20
- Voxtral Mini Transcribe Realtime 3Full report (中文)2026-10-20
- Voxtral Small Transcribe 3Full report (中文)2026-10-20
- Voxtral TTS 3Full report (中文)2026-10-20
- MiniMax-M3.1-FlashFull report (中文)2026-10-19
Value for money
Capability index ≥30 · 3:1 input/output token mix · USD.
Benchmark reports
Each report retains its own task, metrics, test conditions and dates. Full reports currently open in Chinese; model identifiers and original scores are unchanged.
- Aider Polyglot 代码编辑榜单快照External benchmark · 69 records · 2 metrics · Recorded 2026-10-14中文
- DeepSeek-R1 原始蒸馏模型测评Developer-reported · 10 records · 6 metrics · Recorded 2026-10-14中文
- 全球模型综合能力、价格与发布记录External benchmark · 275 records · 1 metrics · Recorded 2026-10-30中文
- EuroEval 欧洲多语言 · 测试集 · few-shotExternal benchmark · 474 records · 33 metrics · Recorded 2026-10-28中文
- EuroEval 欧洲多语言 · 测试集 · zero-shotExternal benchmark · 40 records · 33 metrics · Recorded 2026-10-28中文
- EuroEval 欧洲多语言 · 验证集 · few-shotExternal benchmark · 181 records · 33 metrics · Recorded 2026-10-28中文
- EuroEval 欧洲多语言 · 验证集 · zero-shotExternal benchmark · 92 records · 33 metrics · Recorded 2026-10-28中文
- LiveBench 2.0 综合任务表现与运行成本External benchmark · 49 records · 16 metrics · Recorded 2026-10-18中文
- BFCL V4 工具调用、搜索与记忆External benchmark · 109 records · 31 metrics · Recorded 2026-10-19中文
- ARC-AGI-1 · 抽象推理External benchmark · 291 records · 2 metrics · Recorded 2026-10-28中文
- ARC-AGI-2 · 抽象推理External benchmark · 291 records · 2 metrics · Recorded 2026-10-28中文
- ARC-AGI-3 · Provider AdapterExternal benchmark · 42 records · 2 metrics · Recorded 2026-10-28中文
- ARC-AGI-3 · 标准框架External benchmark · 68 records · 2 metrics · Recorded 2026-10-28中文
- CompassBench 2026-07 · 代码External benchmark · 15 records · 2 metrics · Recorded 2026-10-19中文
- CompassBench 2026-07 · 知识External benchmark · 15 records · 6 metrics · Recorded 2026-10-19中文
- CompassBench 2026-07 · 数学External benchmark · 15 records · 3 metrics · Recorded 2026-10-19中文
- CompassBench 2026-07 · 综合External benchmark · 15 records · 5 metrics · Recorded 2026-10-19中文
- CompassBench 2026-07 · 推理External benchmark · 15 records · 3 metrics · Recorded 2026-10-19中文
- LiveCodeBench 代码生成 · 2024-08-01 至 2025-05-01External benchmark · 28 records · 4 metrics · Recorded 2026-10-19中文
- LiveBench 2.0 综合任务表现与运行成本External benchmark · 32 records · 16 metrics · Recorded 2026-10-31中文
Value rankings require a capability index of at least 30 and valid USD prices. The standard text mix uses three parts input and one part output. Estimated scores, missing prices and retired configurations cannot lead the rankings. Dates for model releases, tests and catalog updates have distinct meanings. Sponsored placements do not change evaluation ranks.