whatmodel.sh / Jais 2 70B Chat
Jais 2 70B Chat
↗ share
+ contribute
Jais 2 70B Chat is an open-weights language model from Inception (G42), released December 2025. Its published size is 72.04B parameters. Licence: apache-2.0.
Catalogue identity notes
Maker repo inception42/Jais-2-70B-Chat @8bcbd3a (requested as inceptionai/Jais-2-70B-Chat, which the HF API redirects to inception42/Jais-2-70B-Chat): licence tag apache-2.0. 72,039,704,576 parameters from the Hugging Face safetensors metadata (dense, so active = total). Released 2025-12: month the maker's Hugging Face repo was created (createdAt 2025-12-09). Comparison-only here: no run command was verified for this checkpoint (architecture Jais2ForCausalLM, model_type 'jais2', ships no custom code; transformers support for it was not verified here). Gated repository (licence acceptance required on Hugging Face).
$ how it compares
The LLMs ranked either side of Jais 2 70B Chat, out of 1132 with a score, then more from Inception (G42).
| # | model | score | where | ||
|---|---|---|---|---|---|
| 321 |  Mercury 2 | 59±2 | comparison only | ||
| 322 |  GLM-4-Plus (0111) | 59±3 | comparison only | ||
| 323 | Jais 2 70B Chat | 59±4 | comparison only | ||
| 324 |  DiffusionGemma 26B-A4B | 59±1 | comparison only | ||
| 325 |  gpt-oss-20b | 58±1 | api $0.13/M +local | ||
| more from Inception (G42) | |||||
| 537 | Jais 2 8B Chat | 51±5 | comparison only | ||
| 713 | Jais Family 30B 16K Chat | 46±3 | local | ||
| 727 | Jais Family 30B 8K Chat | 45±3 | local | ||
$ what it's good at
other tests1 not in a task · 92% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| AraGen-v3 3C3H score (Inception, Claude Sonnet 4 judge) lab | 52.4% ✎ | ▁▁·▁▂▂▄█▄▅▄▄ beats 49/66 | 92% |
$ where to run it
No verified API or download recorded yet · add one
Official weights and documentation: huggingface.co/inception42 ↗
$ how the score was worked out · Inception Arabic/Hindi leader… 92%
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
| source | results | share of the score |
|---|---|---|
 Inception Arabic/Hindi leaderboards (AraGen-v3, HindiGen-v1, Arabic IFEval) · lab-reported | 1 | 92% |
| prior · pull toward the middle | 8% | |
| weighted average | = 58.5 |
$ improve this page
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