Open-weights LLM ranking — coding

scored as 25% aa_coding · 25% aider_polyglot · 15% arena_elo_code · 35% swe_bench_verified (renormalized when signals are missing) — methodology

# model score org params ctx elo $/M out prov
1 glm-5.2 100.0 zai-org 753B 1024k 1469 $3.00 28
2 kimi-k2-instruct 76.1 moonshotai 1000B (A32B) 128k $2.30 1
3 glm-4.6 75.9 zai-org 355B (A32B) 198k $1.74 5
4 kimi-k2.6 72.8 moonshotai 1059B (A32B) 256k $3.20 21
5 glm-5.1 72.4 zai-org 754B 198k 1472 $3.04 19
6 deepseek-v3.2 62.3 deepseek-ai 671B (A37B) 160k $0.34 17
7 minimax-m3 51.7 minimax 427B 1024k $1.20 7
8 deepseek-v3 29.0 deepseek-ai 671B (A37B) 128k $0.80 3
9 gpt-oss-120b 22.5 openai 117B (A5.1B) 128k $0.14 20
10 qwen2.5-coder-32b 20.1 qwen 32.8B 125k $1.00 1
11 llama-4-maverick 13.2 meta-llama 400B (A17B) 1024k $0.60 5
12 gemma-3-27b-it 3.4 google 27B 128k $0.16 5
kimi-k2-0905 100.0 moonshotai 1000B (A32B) 256k $2.50 2
kimi-k2.7-code 93.1 moonshotai 1059B (A32B) 256k $3.15 13
qwen3-coder-480b-a35b 92.9 qwen 480B (A35B) 1024k $1.00 6
deepseek-r1-0528 90.9 deepseek-ai 671B (A37B) 160k $2.15 4
mimo-v2.5-pro 89.7 xiaomi 1023B 1024k 1466 $0.87 5
deepseek-v4-pro 86.2 deepseek-ai 862B 1024k $0.87 16
qwen3-235b-a22b 81.8 qwen 235B (A22B) 128k $1.82 1
deepseek-v4-flash 79.3 deepseek-ai 158B 1024k $0.18 16
qwen3.6-27b 72.4 qwen 27.8B 256k $2.00 9
glm-4.5 71.4 zai-org 355B (A32B) 128k $2.20 1
minimax-m2.7 69.0 minimax 229B (A10B) 200k $0.72 14
qwen3.5-397b-a17b 65.5 qwen 403B (A17B) 250k $2.34 12
kimi-k2-thinking 64.3 moonshotai 1000B (A32B) 256k $2.50 2
deepseek-r1 63.6 deepseek-ai 671B (A37B) 160k $2.50 2
qwen3.5-122b-a10b 58.6 qwen 125B (A10B) 256k $2.08 4
minimax-m2 57.1 minimax 229B (A10B) 200k $1.02 3
glm-4.7 55.2 zai-org 358B 198k $1.75 9
qwen3-coder-30b-a3b 50.0 qwen 30.5B (A3.3B) 160k $0.27 5
gemma-4-31b-it 48.3 google 32.7B 256k $0.35 13
deepseek-v3.1-terminus 44.8 deepseek-ai 671B (A37B) 160k $0.95 5
qwen3.6-35b-a3b 41.4 qwen 36B (A3.3B) 256k $0.97 5
gemma-4-26b-a4b-it 37.9 google 26.5B (A4B) 256k $0.30 10
qwen3-32b 36.4 qwen 32.8B 128k $0.28 5
devstral-small-2512 35.7 mistralai 24B
qwen3-coder-next 34.5 qwen 79.7B 256k $0.80 5
devstral-2-2512 28.6 mistralai 123B 256k $2.00 1
qwen3.5-9b 27.6 qwen 9.7B 256k $0.15 4
qwq-32b 27.3 qwen 32.8B
gpt-oss-20b 20.7 openai 21B (A3.6B) 128k $0.13 13
mistral-small-3.2-24b 13.8 mistralai 24B 125k $0.20 4
llama-3.3-70b-instruct 10.3 meta-llama 70.6B 128k $0.32 12
devstral-small-2507 7.1 mistralai 24B
llama-4-scout 3.4 meta-llama 109B (A17B) 10000k $0.30 4
gemma-3-12b-it google 12B 128k $0.15 1
deepseek-v3.1 deepseek-ai 671B (A37B) 160k $0.79 8
gemma-4-12b-it google 12B
glm-4.5-air zai-org 106B (A12B) 128k $0.85 3
glm-4.6v zai-org 107.7B 128k $0.90 2
glm-4.7-flash zai-org 31.2B 198k $0.40 4
glm-5 zai-org 754B 198k $1.92 14
kimi-k2.5 moonshotai 1059B (A32B) 256k $1.90 11
leanstral-1.5-119b-a6b mistralai 119B (A6B)
llama-3.1-405b-instruct meta-llama 405B
longcat-2.0 meituan-longcat 1776B
magistral-small-2506 mistralai 24B
minimax-m1 minimax 456B (A45.9B) 1000k $2.20 2
minimax-m2.1 minimax 229B (A10B) 200k $1.20 3
minimax-m2.5 minimax 229B (A10B) 200k $0.48 17
mistral-large-2411 mistralai 123B
olmo-2-32b-instruct allenai 32B
olmo-3-32b allenai 32.2B
ornith-1.0-35b deepreinforce-ai 35B
ornith-1.0-397b deepreinforce-ai 397B
ornith-1.0-9b deepreinforce-ai 9B
qwen-agentworld-35b-a3b qwen 34.7B (A3.3B)
qwen2.5-72b-instruct qwen 72.7B 128k $0.40 2
qwen2.5-vl-72b-instruct qwen 73.4B 128k $1.00 1
qwen3-30b-a3b qwen 30.5B (A3.3B) 128k $0.50 3
qwen3-next-80b-a3b-thinking qwen 81.3B (A3B) 256k $0.78 2
qwen3-vl-235b-a22b-instruct qwen 235.7B (A22B) 256k $0.88 5
qwen3-vl-235b-a22b-thinking qwen 235.7B (A22B) 128k $2.60 2
qwen3-vl-30b-a3b-instruct qwen 30.5B (A3B) 256k $0.52 5
qwen3-vl-30b-a3b-thinking qwen 30.5B (A3B) 128k $1.00 2
qwen3-vl-32b-instruct qwen 33.4B 256k $0.42 1
qwen3-vl-4b-instruct qwen 4.4B
qwen3-vl-8b-instruct qwen 8.8B 250k $0.46 2
qwen3-vl-8b-thinking qwen 8.8B 250k $1.36 1
qwen3.5-27b qwen 27.8B 256k $1.56 6
qwen3.5-35b-a3b qwen 36B (A3B) 256k $1.00 9