DeepInfra prices 100+ open models per token as of 2026-10-13: DeepSeek V4 Flash $0.06/$0.18 per million, Kimi K3 $2.85/$14.25, Claude and Gemini also listed.
A long GPU rack aisle in a colocation hall, dense H100-class servers breathing cool air under dim overhead light with one amber status LED glowing — the commodity-hardware economics behind DeepInfra's per-token pricing. Illustration: DeAI
Security X accounts are amplifying Lumen's PoeLLM findings: 3,400+ servers running LiteLLM and Ollama hijacked since April, with C2 hidden in a GitHub poem.
Reuters and Bloomberg corroborate the FT report that Nvidia may deepen its Reflection AI investment or acquire it; Beam's open weights are due this month.
Today in DeAI: White House orders AI incident reporting after Anthropic's disclosures, vLLM takes a Bittensor patch, and LMCache's 9.8 RCE stays unpatched.
Opentensor says Subnet 10 miners hit 54% faster inference, and Pareton's optimization landed in upstream vLLM. The merge checks out; the number stays a claim.
The White House declared AI incident reporting 'not optional' after Anthropic disclosed a fake Philadelphia homicide tip and 20 visa applications.
Mistral put a date on open weights, Ecosia stopped waiting, four hyperscalers began retiring hosted endpoints, and a $50 backdoor showed what downloads cost.
Today in DeAI: Qwen cuts its 7B image model to 8 steps, Anthropic ships a free OSS vulnerability scanner, and OpenAI reports a revenue correction.
Qwen's Qwen-Image-2.1-Turbo runs the same 7B image model in 8 denoising steps instead of 40, with the schedule baked in and the same research-only licence.
Explore by topic
Sponsor disclosure — not editorial
Powered by Morpheus and StrandCMS. Morpheus is a decentralized inference marketplace, covered on the same terms as every other provider. StrandCMS is the open-source, agent-first framework this site is built on.
Learn more →