Udgam Labs releases Vemana, a 278M Telugu base model trained for about $69
Udgam Labs released Vemana, a 278M-parameter Telugu base model with its own 32,000-token tokenizer, trained from scratch on one H100 in about 16 hours for about $69 of GPU time. On 416 Telugu Wikipedia pages created in September and October 2026, it scores 0.361 bits per byte, versus 0.383 for nanochat-telugu-560M, which is twice its size. Sarvam-1, about 9 times larger, is level with or better than Vemana. On multiple-choice Telugu tests Vemana is mostly at chance, and the model makes factual errors and loses coherence over long text.
Open original source udgamlabs.comOne primary source; other links are context, not independent corroboration.
- Primary sourceX discovery · automated brief
- Discovery linkX
- Linked in sourceGGUF files
- Linked in sourceudgamlabs.com/live/vemana
Grok Bot submitted the link. AINewsIQ fetched the primary source and prepared a source-based brief automatically. An editor has not reviewed the story; social attention and the brief do not independently verify its claims.
- Linked article+30
- Fetched source page+30
- No editor link review0
- Original non-discussion link+5
This measures the stored source trail and editorial steps. It does not certify accuracy, peer review, or independent corroboration. Popularity adds no points.
View full scoring →Read the scoring rubric →What the source says
Udgam Labs released Vemana, a 278M-parameter Telugu base model with its own 32,000-token tokenizer, trained from scratch on one H100 in about 16 hours for about $69 of GPU time. On 416 Telugu Wikipedia pages created in September and October 2026, it scores 0.361 bits per byte, versus 0.383 for nanochat-telugu-560M, which is twice its size. Sarvam-1, about 9 times larger, is level with or better than Vemana. On multiple-choice Telugu tests Vemana is mostly at chance, and the model makes factual errors and loses coherence over long text.
A READER'S EVIDENCE CHECKHow to read this signalSix questions about claims, sources, risks, and open checks
This is a signal-level guide until claim-level evidence is available. “Not recorded” means the evidence has not been assessed here; it does not mean a result is false. How we check AI news →
- 01Who is making each claim?
The claimant and whether each claim was measured or asserted are not recorded in this public brief.
- 02What was actually tested?
Not recorded per claim. Check the source for sample size, tasks, model and benchmark versions, and reproducibility.
- 03Where are the sources?
Original link: udgamlabs.com. Brief basis: fetched source page. Primary page date and quote locations are not established in this panel.
- 04What could mislead?
Benchmark numbers were measured by the authors on their own page sets and scorer; source claims not independently checked.
- 05What is at risk if wrong?
Not assessed per claim. Consider safety, privacy, cost, and adoption decisions before relying on the result.
- 06What would change the view?
No claim-specific check is recorded. An inspectable primary method or independent replication could change the assessment.
Key findings from the stored brief
- Vemana scores 0.361 bits per byte on 416 fresh Telugu pages, versus 0.383 for nanochat-telugu-560M.
- The lead over nanochat-telugu-560M holds on the other two page sets and at a 512-token window.
- Sarvam-1 at 2.5B parameters is level with Vemana on two page sets and better on the July–August set.
- On Belebele-te, MILU-te and IndicXNLI-te, Vemana scores close to chance accuracy.
Benchmark numbers were measured by the authors on their own page sets and scorer; source claims not independently checked.
This is a source-based abstract, not independent validation of the source's claims.
Evidence & sources
These links show where the signal came from and what the source page linked to. A linked page is context, not independent corroboration.
These are Grok Bot discovery notes. AINewsIQ prepared a brief from available source material. The bot's rationale, linked posts, authors, and stances have not been independently verified.
Udgam Labs' Jitesh Prasad released Vemana (Apache-2.0), a 278M-parameter Telugu base model with its own tokenizer, trained in public in about 16 hours on one H100. On Telugu Wikipedia pages from Sep–Oct 2026 it has the lowest bits-per-byte among open base models under 1B they tested; the page notes Sarvam-1 (2.5B) is level or better. Runs in-browser. Self-reported.
Benchmark numbers were measured by the authors on their own page sets and scorer; source claims not independently checked.
Engagement counts, when shown, are observed snapshots. They do not establish accuracy or corroboration.
- x post ↗ · 1,046 likes, 106 reposts, 42 replies, 11 quotes · observed 10/27/2026, 4:11:28 PM
- huggingface post ↗ · 3 likes, 2 downloads · observed 10/27/2026, 4:10:09 PM
Related archive coverage
These relationships record a citation or topic overlap, not confirmation of the finding.
Grok Bot submitted the link. AINewsIQ fetched the primary source and prepared a source-based brief automatically. An editor has not reviewed the story; social attention and the brief do not independently verify its claims.
Timeline & updates
- Signal dated by source or discovery feed
This timestamp places the signal in the radar. It may be the discovery post date rather than the primary page's first publication date.
- Shared brief stored
AINewsIQ prepared one abstract from fetched source page.
A publication revision history is not stored for this signal. New engagement snapshots do not mean the source itself changed.
Why this may matter
The release reports a low-cost, openly documented training run for a small Telugu model and a bits-per-byte comparison on post-release pages. It gives a concrete reference point for small non-English base models.
Limits and open questions
Benchmark numbers were measured by the authors on their own page sets and scorer; source claims not independently checked.
Organization-specific financial, security, and operational effects require the organization's own context and review.
Discussion
Comments are reviewed before publication. Discuss the evidence, not the person.
No published comments yet.
Reader discussion opens with accounts. You can read the original source and published comments now.