Context
524K tokens (~262 books)
Input $/1M
n/a
Output $/1M
n/a
Type
text
License
Open Source
Benchmarks
0 tested
Data as of
About
Jais 3 36B-A4B combines MoE layers with MoVA attention, with 36B total and about 4B active parameters and a 512K token context window. Part of Jais 3, the Arabic-centric agentic model family from MBZUAI Institute of Foundation Models with Inception, built on K2 Horizon. Weights, training checkpoints and code under Apache 2.0.
No benchmark data available yet.
Links
Research
Documentation
Community
Source Code
BenchGecko API
jais-3-mova-36b-a4b
Specifications
- Typetext
- Context524K tokens (~262 books)
- ReleasedOct 2026
- LicenseOpen Source
- StatusActive
Available On
M
MBZUAIn/aShare & Export
Frequently Asked Questions
Jais 3 36B-A4B is an open-source text AI model by MBZUAI, released in October 2026. Context window: 524K tokens.
Related Models
Qwen3 Coder Next FP8 · AlibabaQwen2.5 Coder 32B Instruct AWQ · AlibabaSpeaker Diarization Community 1 · PyannoteJais 3 0.9B · MBZUAIGemma 3 12B (free) · Google DeepMindBenchmarks
Key facts · as of 2026-10-16
- Jais 3 36B-A4B by MBZUAI. Not enough public benchmark scores to rank yet.
- List price n/a input · n/a output per 1M tokens (as of 2026-10-16).
How to cite · data as of 2026-10-16
Jais 3 36B-A4B · benchmarks, pricing and providers. BenchGecko, data as of 2026-10-16. https://benchgecko.ai/model/jais-3-mova-36b-a4b
Credit "Source: BenchGecko" with a link. Prices per provider and Gecko Tests are BenchGecko data (CC BY 4.0); benchmark scores keep their original source, listed in the JSON. JSON · llms.txt · MCP