Skip to content

Newest models

Most recent releases, with whatever independent scores exist so far.

Most recent releases, with whatever independent scores exist so far. Source: OpenRouter / Epoch AI. Variants (reasoning effort, batch) are folded into one row.

#ModelCapabilityReasoningCodingContext$/M
51Hy3 previewTencent———262K$0.28
52MiMo-V2.5Xiaomi———1.05M$0.17
53MiMo-V2.5-ProXiaomi———1.05M$0.54
54Kimi K2.6Moonshot AI151.090.8—262K$1.34
55Qwen 3.6 35B-A3BAlibaba (Qwen)143.9————
56GLM 5.1Z.ai (Zhipu AI)149.989.9—205K$2.15
57Gemma 4 31BGoogle142.775.8—262K$0.15
58Gemma 4 26B A4BGoogle141.873.2—262K$0.14
59Trinity Large ThinkingArcee AI———262K$0.39
60Reka EdgeReka AI———16K$0.10
61MiniMax M2.7MiniMax145.8——205K$0.37
62Mistral Small 4Mistral AI · +1 variant———262K$0.26
63Nemotron 3 SuperNVIDIA———262K$0.17
64Qwen3.5-122B-A10BAlibaba (Qwen)———262K$0.71
65Qwen3.5-27BAlibaba (Qwen)———262K$0.54
66Qwen3.5-35B-A3BAlibaba (Qwen)142.583.5—262K$0.45
67Qwen3.5-9BAlibaba (Qwen)139.479.0—262K$0.11
68Qwen3.5-2BAlibaba (Qwen)—————
69Qwen3.5-4BAlibaba (Qwen)—————
70Qwen3.5 397B A17BAlibaba (Qwen)146.686.4—262K$1.29
71MiniMax M2.5MiniMax146.7——205K$0.47
72GLM 5Z.ai (Zhipu AI)145.887.8—205K$0.93
73Qwen3 Coder NextAlibaba (Qwen)———262K$0.29
74Step 3.5 FlashStepFun———262K$0.15
75Kimi K2.5Moonshot AI148.1——262K$0.90
76Kimi K2.5 (Fireworks)Moonshot AI—87.6———
77GLM 4.7 FlashZ.ai (Zhipu AI)—60.5—200K$0.15
78MiniMax M2.1MiniMax———205K$0.53
79GLM 4.7Z.ai (Zhipu AI)143.583.3—205K$1.00
80GLM-4.7 (Novita)Z.ai (Zhipu AI)—————
81GLM-4.7 (Together)Z.ai (Zhipu AI)—————
82Nemotron 3 Nano 30B A3BNVIDIA———262K$0.087
83Devstral 2 2512Mistral AI———262K$0.80
84GLM 4.6VZ.ai (Zhipu AI)———131K$0.45
85Ministral 3 14B 2512Mistral AI———262K$0.20
86Ministral 3 3B 2512Mistral AI———131K$0.10
87Ministral 3 8B 2512Mistral AI · +1 variant———262K$0.15
88Mistral Large 3Mistral AI—————
89DeepSeek V3.2DeepSeek146.383.4—164K$0.32
90DeepSeek-V3.2 (Thinking; Fireworks)DeepSeek—————
91DeepSeek-V3.2 (Thinking; Novita)DeepSeek—————
92DeepSeek-V3.2-SpecialeDeepSeek—————
93Kimi K2 ThinkingMoonshot AI146.0——262K$1.07
94Kimi K2 Thinking (Together)Moonshot AI—————
95Kimi K2 Thinking TurboMoonshot AI—84.2———
96Voxtral Small 24B 2507Mistral AI———33K$0.15
97gpt-oss-safeguard-20bOpenAI———131K$0.13
98MiniMax M2MiniMax———205K$0.53
99Qwen3 VL 32B InstructAlibaba (Qwen)———131K$0.18
100Granite 4.0 MicroIBM—28.3—131K$0.041
— means no published score from that source yet · click a model for every benchmark with its source
Best AI for coding →
Models ranked on DeepSWE
Best AI for reasoning →
Ranked on GPQA Diamond
Best AI for math →
Ranked on competition math
Best open-weight models →
Weights you can run yourself
Cheapest capable models →
Lowest blended API price
Best AI coding tools →
Apps & agents by popularity
Best AI image generators →
Apps & models
Best AI for research →
Answer engines & deep research