AI War Tracker
298 models in catalog

AI models

Every model we track — frontier flagships, open-weights specialists, narrow benchmarks-only releases. Filter by lab, country, access, modality, or release window. Sorted by newest by default; head to the leaderboard for the ranked view.

Leaderboard →Labs →Benchmarks →

Updated May 29, 2026 · Benchmarks via Artificial Analysis, specs & pricing via OpenRouter · Methodology · Spotted an error?

RankModelCoding idxSciCodeAider Polyglot EditMultiPL-EMBPPSWE-bench ProAider PolyglotLiveCodeBenchTerminal-BenchSWE-bench VerifiedHumanEvalReleasedCountryTypeAccessParamsCutoffContextSpeedLatencyIn $/MOut $/M
Microsoft17.62623.13.882.82025llmOpen weights202416K330.20$0.07$0.14
Meta17.11767.832.81.52025multimodalOpen weights109B total / 17B active (MoE)202410M7760.31$0.08$0.30
Alibaba16.925.58.32026llm2350.99$0.10$0.80
Alibaba16.917.1321.52025llm$0.00$0.00
Amazon16.720.823.36.1892024multimodalAPI only300K1000.50$0.80$3.20
Sarvam16.517.829.52.32025llm1361.17$0.00$0.00
Google16.424.48.32026llm$0.00$0.00
Upstage16.224.77.62026llmAPI only128K$0.15$0.60
IBM16.120.925.12.32025llm5248.71$0.10$0.30
OpenAI15.725.96.29.832.63.82025multimodalAPI only20241M2002.00$0.10$0.40
NVIDIA14.923.316.94.52024llmOpen weights7000000000020232920.24$1.20$1.20
Upstage14.626.92.32025llm$0.00$0.00
Alibaba14.213.72902025llm$0.00$0.00
IBM14.125.82.32026llm$0.00$0.00
Sarvam1426.41.52026llm1281.29$0.00$0.00
Nanbeige13.326.602026llm$0.00$0.00
Meta13.218.919.80.82024llmOpen weights20238K450.70$0.51$0.74
AI21 Labs13.118.818.12.32025llmOpen weights2024256K480.97$2.00$8.00
Mistral AI1314.424.702025multimodalOpen weights131K1540.34$0.10$0.10
Google12.921.213.73.82025llm$0.10$0.30
Alibaba12.66.930.802025llm1380.97$0.10$0.40
Allen Institute for AI12.310.326.602025llm$0.10$0.20
Google1220.932026llm$0.00$0.00
NVIDIA11.616.46.82026llm$0.00$0.00
Anthropic11.618.615.40.875.92024multimodalAPI only2023200K1040.40$0.25$1.25

Ranked on Coding. Cell colors show relative standing within each column (red → yellow → green). Scores are curated approximations — see each model for sources.