Sarvam open-sources its latest 105B model; founder says scaling agentic RL 'is not for the faint hearted'
itsOmSarraf_ · x · 2026-11-23
Indian AI company Sarvam announced it is open-sourcing the latest edition of Sarvam-105B. The founder's tl;dr: scaling agentic RL is 'not for the faint hearted' — data generation, evals, and hill climbing in non-verifiable domains is a massive unlock, but much remains to be done. Next stop: 1T parameters.
More from Models
- 55-Page Riemann Hypothesis Proof Guide, De-Formalized from Lean with Gemini and GPT — chaumian · 2026-11-23
- Gemini 4 Neon xhigh costs 8.6x more but adds just 1-4 points on most benchmarks — jeremyphoward · 2026-11-23
- Poland's 9B-parameter Thalamus Reasoning tops LongevityBench, beating 29 models — Dr_Singularity · 2026-11-23
- Dev: Gemini 4 Neon rewrote my training kernels and supervised runs; rumors point to bigger models — xennygrimmato_ · 2026-11-23
- Ant Group open-sources LingSante-3.1-air-VL, a 124B MoE medical multimodal model — jacek2023 · 2026-11-23
- Inkling 2 Mini: a 26B local model that runs on laptops, and someone rebuilt Apple Notes around it — simonguozirui · 2026-11-23