The open market for production LLM inference
Bucharest
Joined October 2026
- Tweets66
- Following23
- Followers5
- Likes20
Pinned Tweet
llm.buzz is the open market for production inference.
Providers compete on price and speed to give you the best prices on the market.
Building with models: one API, every model, the cheapest healthy provider on each request.
Serving models: list your endpoint, set your price, keep 100%. Your own model too.
84 models on llm.buzz, and any of them can take another seller.
If you serve one for less than the price you see, list your endpoint next to it.
Apply once and you set the price. An automated check sends real requests, and the listing goes live when it passes.
We take 0%.
Open weights with no third-party host yet:
Mistral Large 4 from @MistralAI: 1.05T total, 52B active, 1M context.
Flint from @Recursive_SI: 40B total, 2.5B active, Apache 2.0.
Serving one? List it on llm.buzz. The first seller sets the price.
Claude Sonnet 5 on llm.buzz: $1.60 in, $8 out per million tokens.
@AnthropicAI list: $2 in, $10 out.
Cache reads $0.16, 20% off as an Auto deal.
Prices as of 10:16 UTC today.
Mistral Large 4 went open weights today. 1.05T total, 52B active, 1M context.
The release from @MistralAI takes text and image input.
If you're serving it, list it on llm.buzz. You set the price, an automated check runs, and it goes live when it passes. We take 0%.
Opus 5.5 is $3/$15 on llm.buzz. Sonnet 5.5 is $2/$10. DeepSeek V4.1 Flash is $0.15/$0.45. Cache reads from half a cent.
85 models live on llm.buzz right now.
Opus 5.5 is $3 in, $15 out. Claude 3.5 Sonnet v2 is $1.50 in, $7.50 out.If you are serving a model, list it today. You set the price, an automated check runs, and it goes live when it passes. We take 0%.
EU-only routing comes with every plan on llm.buzz.
Name EU providers in a request and your prompt stays with them, failover included.
Surogate 3.7 Flash is only on llm.buzz.
An independent EU provider serves it at $0.10 in, $0.75 out per million. They set that price, and they keep all of it.
Trained or fine-tuned a model of your own? You can sell it.
Serve it from any OpenAI-compatible endpoint, list it on llm.buzz at your price, and every caller on the market can use it. We approve a new model once, then it's yours to price.
Surogate 3.7 Flash went live this way, from an independent EU provider.
