Open-Weight AI Models

Open weights buy you three things a closed API cannot: the option to self-host, freedom from per-token pricing at scale, and no risk of the model being deprecated out from under you. The listed prices are what hosted providers charge to run these for you; the real economics show up only when you run them yourself. Good fit for high-volume, latency-sensitive, or data-residency-constrained workloads.

Showing top 25 of 369 matching models · last synced 15 hours ago

Top pick right now: Aya Vision 8B from Cohere. It is the lowest-priced open-weight option here at N/A/Mtok output, and you can self-host it to drop that cost further.

# Model Provider Input $/Mtok Output $/Mtok Context Capabilities
5
Whisper
Groq N/A N/A N/A open
6
Whisper Large V3 Turbo
Groq N/A N/A N/A open
15
LFM2.5-1.2B-Instruct (free)
OpenRouter Free Free 32,768 structured output, open
16
LFM2.5-1.2B-Thinking (free)
OpenRouter Free Free 32,768 tools, reasoning, structured output, open
17
Trinity Large Preview
OpenRouter Free Free 131,072 tools, structured output, open
18
Uncensored (free)
OpenRouter Free Free 32,768 structured output, open
19
Seedream 4.5
OpenRouter Free Free 4,096 vision, open
20
FLUX.2 Klein 4B
OpenRouter Free Free 40,960 vision, open
21
Hermes 3 405B Instruct (free)
OpenRouter Free Free 131,072 open
22
Llama 3.2 3B Instruct (free)
OpenRouter Free Free 131,072 open
23
Llama 3.3 70B Instruct (free)
OpenRouter Free Free 131,072 tools, open
24
Laguna M.1 (free)
OpenRouter Free Free 262,144 tools, reasoning, open
25
Laguna XS.2 (free)
OpenRouter Free Free 262,144 tools, reasoning, open

Browse other categories

Data synced daily from models.dev. Always verify pricing with the provider before deploying to production.