ALLaM-2-7b
- Released
- 1y ago
- Size
- 406Bparameters
Llama 3.1 405B post-trained with the Tulu 3 recipe, the largest model trained with RL with verifiable rewards at release.
How it compares
Frequently asked questions
Running locally
Tulu 3 405B needs an estimated 283.5 GB of VRAM at Q4, 496.4 GB at Q8 and 934 GB at FP16. All figures are estimates, covering the weights plus about 15% and 0.5 GB for a short context.
No, Tulu 3 405B does not fit on a 24 GB, 16 GB, 12 GB or 8 GB GPU at Q4, Q8 or FP16, according to estimates.
Yes, with sufficiently large unified memory. Assuming macOS lets the GPU use about 75% of unified memory and the estimated 283.5 GB needed at Q4, the smallest common Mac memory size that fits is 512 GB (estimate).
Privacy and data
Yes, Tulu 3 405B has open weights, so you can self-host it for full control of your data.
Capabilities and specs
No, Tulu 3 405B only accepts text input.
Quality and benchmarks
The closest alternatives to Tulu 3 405B, matched on model kind and release date with no comparable price, are ALLaM-2-7b by SDAIA, DeepSeek-R1-Distill-Qwen-32B by DeepSeek and Phi-4-mini by Microsoft.
ALLaM-2-7bTulu 3 405B vs ALLaM-2-7bDeepSeek-R1-Distill-Qwen-32BTulu 3 405B vs DeepSeek-R1-Distill-Qwen-32BPhi-4-miniTulu 3 405B vs Phi-4-mini
Licensing and access
Tulu 3 405B has open weights and is released under the Llama 3.1 Community License.
The Tulu 3 405B weights can be downloaded from the Hugging Face repository allenai/Llama-3.1-Tulu-3-405B.
Basics
Tulu 3 405B was made by Ai2 and released on January 30, 2025.
Yes, Tulu 3 405B has open weights, so you can download and run it yourself.
Generated from Vertical Tabs data, updated November 3, 2026.
Discussion
No comments yet
Be the first to share your thoughts.