PulseAugur
EN
LIVE 03:50:19

Sarvam 30B and 105B models fail on Indian language tasks, Qwen 3.8 27B succeeds

By PulseAugur Editorial · [1 sources] ·

A user attempted to use Sarvam 30B and 105B models for Indian language tasks such as translation and entity extraction, but found them to be ineffective. The models frequently repeated tokens and failed to follow instructions, regardless of prompt engineering or configuration adjustments. In contrast, the Qwen 3.8 27B model performed surprisingly well on the same tasks. AI

IMPACT Highlights potential issues with Sarvam models for Indian languages and suggests Qwen as a viable alternative.

RANK_REASON User experience report on model performance.

Read on Mastodon — mastodon.social →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Sarvam 30B and 105B models fail on Indian language tasks, Qwen 3.8 27B succeeds

How we ranked this

Signal score
4 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Commentary
User experience report on model performance.
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
model release, product
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. Mastodon — mastodon.social TIER_1 English(EN) · abhiyantrik ·

    I tried using # sarvam 30b and 105b on Indian language tasks. Nothing fancy, just tried doing translation, entity extraction. I could not make it work, no matte

    I tried using # sarvam 30b and 105b on Indian language tasks. Nothing fancy, just tried doing translation, entity extraction. I could not make it work, no matter how much prompt engineering I tried with or the other configurations. The model started repating tokens most of the ti…

RELATED ENTITIES

RELATED TOPICS