Tulu 3 70B

Estimated VRAM is 49.7 GB at Q4, 86.7 GB at Q8 and 162.8 GB at FP16, derived from the parameter count of 70.6B. All figures are estimates: weights plus about 15% and 0.5 GB for a short context.

No. On a 24 GB, 16 GB, 12 GB or 8 GB GPU it does not fit at Q4, Q8 or FP16 (estimate).

Yes, with enough unified memory. At Q4 the estimated VRAM is 49.7 GB. The estimate assumes macOS lets the GPU use about 75% of unified memory, and the smallest common Mac memory size that fits Q4 is 96 GB (estimate).

Yes. Tulu 3 70B has open weights, so you can self-host it for full control of your data.

Run Tulu 3 70B locally

No. Tulu 3 70B only accepts text as input; it cannot read images, audio, video or PDFs.

Mistral Large 2.1 by Mistral AI, Qwen2.5-Coder-0.5B by Alibaba and Aya Expanse 32B by Cohere, matched on model kind and release date (no comparable price).

Mistral Large 2.1Tulu 3 70B vs Mistral Large 2.1Qwen2.5-Coder-0.5BTulu 3 70B vs Qwen2.5-Coder-0.5BAya Expanse 32BTulu 3 70B vs Aya Expanse 32B

Tulu 3 70B is open weights. It is released under the Llama 3.1 Community License.

The weights are available on the Hugging Face repository allenai/Llama-3.1-Tulu-3-70B.

Run Tulu 3 70B locally

Tulu 3 70B was made by Ai2 and released on November 21, 2024.

More from Ai2

Yes, Tulu 3 70B has open weights.