Tülu 3 70B

Overview

Allen AI's fully open RLHF stack on Llama 3.1 70B, beating Claude Haiku, GPT-3.5 Turbo, and GPT-4o-mini on standard reasoning and code benchmarks.

When to pick this model

• Self-hosted alternative to closed mid-tier APIs
• Math-heavy chat with GSM8K 93.5 territory performance
• Code assistance where HumanEval+ matters more than agentic loops
• Research projects that need a fully documented post-training pipeline
• Workloads that justify a 2x A100 footprint

VRAM requirements by quantization

VRAM figures include model weights plus a typical 8k KV cache and ~600 MB runtime overhead (Ollama / llama.cpp baseline). Add headroom for higher context lengths.

Published benchmark scores

Scores published by the model author or aggregated from public leaderboards. Re-measured monthly by our editorial team.

Strengths

• Beats Claude Haiku, GPT-3.5 Turbo, and GPT-4o-mini on key evals
• GSM8K 93.5 and HumanEval+ 92.4 at open weights
• Fully open SFT + DPO + RLVR recipe
• Strong instruction following and refusal calibration
• Stable, well-documented behavior for production deploys

Limitations

• ~40 GB VRAM at Q4 — needs serious hardware
• Bound by Llama 3.1 Community License
• No multimodal capabilities

Architecture & training

Architecture: Dense Llama 3.1 70B · full Tülu recipe

Training: SFT + DPO + RLVR on 70B.

Quick start

ollama run tulu3:70b

Or use the open-source MCP server to query this model from Claude Desktop, Cursor, or any MCP-compatible client.