whatmodel.sh / North Small Translate 1.0

North Small Translate 1.0

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

North Small Translate 1.0 is an open-weights language model from Cohere, released August 2026. Its published size is 218B parameters (25B active per token) and a 16k-token context window. Licence: cc-by-nc-4.0.

Catalogue identity notes

Cohere's translation model (API id north-small-translate-1-0; weights CohereLabs/North-Small-Translate-1.0 @e5b9495, created 2026-08-14, licence cc-by-nc-4.0, gated). Subject of arXiv 2609.13916v1, whose Table 1 row 'North Small Translate' is this model (same 218B (25B) size). The repo is gated and only vLLM fp8/w4a16 variants exist from Cohere; no verified local command could be read (card behind the gate), so local is left null. Translation-specialised: its rows are on the three translation tests only.

Model information sources

Release date ↗ · Weights licence ↗

Cohere docs model table: 'North Small Translate is a 218B total / 25B active parameter MoE model purpose-built for machine translation across more than 50 languages'; paper Table 1 '218B (25B)'; HF safetensors metadata 218,254,938,112.

71.4±5.9#71 of 719 LLMsfrom 3 benchmark results across 1 source

$ how it compares

The LLMs ranked either side of North Small Translate 1.0, out of 828 with a score, then more from Cohere.

#modelscorewhereRAMreleased
74Grok 4.20 (0309 reasoning) SpaceXAI (xAI)72±1api $1.56/M2026-04
75Solar Open 2 (250B-A15B) Upstage72±1local~527GB2026-07
76North Small Translate 1.0 Cohere71±6api price unknown2026-08
77GPT-5.4 mini OpenAI71±1api $1.69/M2026-03
78Ling 3.0 Flash (124B-A5B) inclusionAI (Ant Group)71±2api $0.090/M +local~75GB2026-08
more from Cohere
175Command A+ (05-2026) Cohere64±1api price unknown +local~71GB2026-05
277Command A (03-2025) Cohere57±1api $4.38/M +local~37GB2025-03
324Command A Reasoning (08-2025) Cohere55±2local~34GB

see all LLMs · find the ones that fit your machine

$ what it's good at

other tests3 not in a task · 85% of its score
testresultvs the others in its rundifficultycountswhy
WMT25 Terminology Translation (term success rate) lab89.7% ✎▂·▂··▂·▄▂·█▆ beats 7/12medium · avg 64%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 42%noisy
WMT24++ lab80.6% ✎█·▃·▅▅▃···▃▃ beats 9/10easy · avg 71%▇▇▇▇▇▇▇▇▇▇ 24%blunt
WMT26 General MT (GEMBA-ESA, LLM judge) lab83.6% ✎▄·▄▄▄·▄██▄█▄ beats 12/12easy · avg 65%▇▇▇▇▇▇▇▇ 19%blunt

$ where to run it

Through an API

api · price unknown · 16k context · docs.cohere.com ↗

Weights and documentation: huggingface.co/CohereLabs ↗

$ how the score was worked out · Cohere — North Small Translat… 85%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
Cohere — North Small Translate (arXiv:2609.13916v1) · lab-reported3▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 85%CC BY 4.0
prior · pull toward the middle▇▇▇▇▇ 15%
weighted average= 71.4

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json