{"id":"sarvam-prism","slug":"sarvam-prism","name":"Sarvam Prism","orgId":"sarvam-ai","family":"sarvam-prism","releaseDate":null,"availability":"ga","openWeights":true,"license":"Apache 2.0","paramsTotal":null,"paramsSource":null,"contextWindow":null,"knowledgeCutoff":null,"modelCardUrl":"https://x.com/SarvamAI/status/2117621415520448838","hfRepo":null,"modalities":["text"],"aliases":["Sarvam Prism","sarvam prism","sarvam-prism","sarvamai/prism","sarvam ai prism"],"aliasNote":null,"origin":"tracked","tracked":{"trackedSince":"2026-11-03","announced":"2026-11-03","blurb":"Sarvam AI's decision model built on Sarvam 105B: given a context and questions with fixed answer options, it returns a probability for every option in a single forward pass instead of generating text. English, 10 Indian languages and Hinglish. Open weights under Apache 2.0, plus Sarvam's API.","api":"live","apiModelId":"sarvam-prism","openWeights":"yes","params":{"text":"104,991,889,280 stored parameters (Sarvam 105B mixture-of-experts backbone with 128 routed experts per layer, plus a decision head)","source":"Sarvam's Hugging Face safetensors metadata (sarvamai/prism)"},"facts":[{"label":"Type","value":"Decision model: one forward pass with a scoring head returns a probability per answer option; it does not generate text"},{"label":"Base model","value":"LoRA fine-tune of Sarvam 105B"},{"label":"Question types","value":"Choice, yes/no and score (ordered levels)"},{"label":"Training","value":"Cross-entropy with 5% label smoothing plus a Brier loss"},{"label":"Context","value":"131,072 tokens per question with the open weights; the hosted API accepts 32,768"},{"label":"Languages","value":"English, 10 Indian languages (Hindi, Bengali, Tamil, Telugu, Marathi, Gujarati, Kannada, Malayalam, Odia, Punjabi) and Hinglish"},{"label":"Inputs","value":"Text or JSON; no images, and no streaming in the API"},{"label":"Weights","value":"Hugging Face sarvamai/prism, not gated: a custom FP8 layout loaded with Sarvam's prism.py (about 106 GB); needs one FP8 GPU with about 120 GB free memory, tested on B200"},{"label":"License","value":"Apache 2.0"},{"label":"API","value":"Sarvam API model sarvam-prism at POST /v2/systemone: ₹4 per 1M input tokens, output free"}],"claims":["Lab claim: 71.8 average across three decision indices (Jev Decision Index 67.0, Sarvam's recreation of the Artificial Analysis Decision Index 65.9, an internal Indian-language benchmark 82.5), the highest average of the 25 decision models Sarvam tested that answer in under 100 ms; mammoth-1 (75.2) and GLiDE 1.2 (73.2) score higher but are slower.","Lab claim: 44 ms median per request (95th percentile 420 ms) on one NVIDIA B200.","Lab claim: at 90% or higher confidence, 2.2% of English answers and 4.4% of Indian-language answers are wrong, against 3.3% and 5.4% expected from the stated confidence.","Lab claim: 78.5% on judging agent runs (700 runs) and 89.0% on an internal security benchmark (5,517 questions); 0 of 2,000 answers change when the options are shuffled."],"sources":[{"kind":"lab","title":"Sarvam AI announcement on X","url":"https://x.com/SarvamAI/status/2117621415520448838","date":"2026-11-03"},{"kind":"lab","title":"Sarvam Prism (Sarvam blog)","url":"https://www.sarvam.ai/blogs/sarvam-prism","date":"2026-11-03"},{"kind":"lab","title":"Sarvam docs: Sarvam Prism","url":"https://docs.sarvam.ai/api/token-factory/models/sarvam-prism"},{"kind":"lab","title":"Sarvam API reference: SystemOne decisions","url":"https://docs.sarvam.ai/api-reference/decisions/systemone"},{"kind":"repo","title":"sarvamai/prism on Hugging Face","url":"https://huggingface.co/sarvamai/prism"}],"firstBenchmark":null},"organization":{"id":"sarvam-ai","name":"Sarvam AI","slug":"sarvam-ai","country":null,"url":"https://www.sarvam.ai","color":"#11110F","mark":"S"},"price":null,"answer":{"word":"Pending benchmarks.","scope":"Announced Nov 3, 2026; no independent results yet.","lead":"As of Nov 22, 2026, Sarvam Prism is pending benchmarks on isitsota: announced Nov 3, 2026, with no independent results on the boards we track yet.","confidence":"low"},"status":"PENDING_BENCHMARKS","verdicts":[]}