VLM Run launches TypeSafe-compatible API: typed visual Q&A with calibrated probabilities at ~44ms p50
spillai · x · 2026-09-26
VLM Run released a TypeSafe-compatible API (System One) that answers typed questions about text, JSON, or images and returns calibrated probabilities for every answer—no generation, no parsing, so answers can never go off-schema.
- Wire-compatible with TypeSafe's Jev API: point typesafe-sdk at https://gateway.vlm.run/typesafe and it works unchanged
- Fast: one forward pass instead of decode-and-parse, p50 44ms, deliberately held under 200ms
- Cheap: a fraction of a cent per read, suitable for gating expensive pipelines, document routing, frame labeling, upload sanity checks
- No fine-tuning: engines are open-weight VLMs, so editing criteria suffices for new questions—no dataset or redeploy
- Initial models: gemma4-26B-A4B-it, diffusiongemma-26B-A4B-it, Qwen3.8-27B, Qwen3.5-0.8B
More from coding & agent
- Claude Opus 5.5 generates a full launch video — animation, music, voiceover — in 20 minutes — cedric_chee · 2026-09-26
- OpenRouter launches typesafe/jev-router, a cache-aware router that picks models per request — alexcovo_eth · 2026-09-26
- Data labeling firm predicts enterprise evals will become core IP as revenue shifts to Fortune 1000 — soumitrashukla9 · 2026-09-26
- asc builds upload now prints build IDs and surfaces Apple's real errors — rudrank · 2026-09-26
- asc CLI pre-checks emojis locally, naming each character Apple would reject — rudrank · 2026-09-26
- asc CLI's --if-exists now covers pricing, metadata push, and capabilities — rudrank · 2026-09-26