whatmodel.sh / Tulu 2 DPO 70B
Tulu 2 DPO 70B
↗ share
+ contribute
Tulu 2 DPO 70B is an open-weights language model from Ai2, released November 2023. Its published size is 69B parameters. Licence: ai2-impact-low-risk.
Catalogue identity notes
Comparison-only. Arena entry 'tulu-2-dpo-70b' (text). LMArena's leaderboard table (arena-leaderboard space @6f2eeeb) labels 'tulu-2-dpo-70b' as 'Tulu-2-DPO-70B' (AllenAI/UW, AI2 ImpACT Low-risk), linking https://huggingface.co/allenai/tulu-2-dpo-70b. Maker repo allenai/tulu-2-dpo-70b @0ab5c87: licence tag other (ai2-impact-license-low-risk). 68,976,648,192 parameters from the Hugging Face safetensors metadata (dense, so active = total). Released 2023-11: month the maker's Hugging Face repo was created (createdAt 2023-11-12).
$ how it compares
The LLMs ranked either side of Tulu 2 DPO 70B, out of 828 with a score, then more from Ai2.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 584 |  CodeGemma 7B IT (v1.0) | 43±1 | local | ||
| 585 |  Aya Vision 32B | 42±3 | api price unknown +local | ||
| 586 | Tulu 2 DPO 70B | 42±1 | local | ||
| 587 |  Mathstral 7B v0.1 | 42±1 | local | ||
| 588 |  Yi-1.5 9B (base) | 42±2 | local | ||
| more from Ai2 | |||||
| 259 | Olmo 3.1 32B Think | 58±1 | local | ||
| 262 | Olmo 3 32B Think | 58±1 | local | ||
| 314 | Molmo 2 8B | 55±2 | local | ||
$ what it's good at
coding▁▁▁▃▆██▇▆▄▃▁ #377 of 458 * · 5.7% of its score · 1 test
Beats LFM2.5 8B-A1B, DeepSeek R1 Distill Qwen 7B, Qwen3.5 2B here, though they rank higher overall. Trails Qwen2.5 1.5B Instruct, WizardLM 13B V1.2, Falcon3 3B Instruct here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Tulu 2: CodexEval pass@10 lab | 68.9% ✎ | ▃▃▃·▆▃▃█▃▂▂▅ beats 16/25 | 5.7% |
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #317 of 393 · 59% of its score · 5 tests
Beats Gemma 4 E4B, OpenThinker3 7B, Gemma 4 E2B here, though they rank higher overall. Trails LFM2 350M, Gemma 3 1B, Flan-T5 XXL here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| AlpacaEval 2 length-controlled | 21.2% ✎ | ▁▂▂▃▆▆█▆▄▄▁▁ beats 55/122 | 24% |
| Arena-Hard (v0.1, GPT-4-1106 judge) | 15.0% ✎ | ▂▁▁▅█▅▄▃▃▅▃▁ beats 8/60 | 16% |
| Polyrating rating (bias-aware re-fit of Chatbot Arena votes, mid-2024) | 1122.0 ✎ | ▂▁▂▁▃▇█▆▇▄▂▂ beats 57/95 | 8.2% |
| LMArena (text) | 1177.6 ✎ | ▁▁▁▂▃▃▃▆▆▇█▃ beats 54/379 | 8.1% |
| AlpacaEval 1.0 win rate (vs text-davinci-003, GPT-4 judge) | 95.0% ✎ | ▂▂·▆·▇▃▆█▆▄▂ beats 42/45 | 1.9% |
math▁▂▃▅▆▅▄█▄▁▁▁ #234 of 398 * · 4.4% of its score · 1 test
Beats GPT-4o (2024-08-06), Command A (03-2025), GPT-4o (2024-11-20) here, though they rank higher overall.
other tests5 not in a task · 31% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| BIG-Bench Hard lab | 66.0% ✎ | ▁·····▂█▂▅·▁ beats 21/25 | 14% |
| WildBench v2 WB-Score (task macro) lab | 27.9838 WB-Score ✎ | ▁·▁▁·▃█▄▇▆▅▆ beats 10/52 | 9.2% |
| MMLU lab | 67.8% ✎ | ▁▁▃▃▄█▄▁█··▁ beats 23/25 | 4.9% |
| Open PT-LLM Leaderboard average (9 Portuguese tasks) | 74.1% ✎ | ▃▁▁▁▁▁▁▁▂█▅▁ beats 213/276 | 1.8% |
| Tulu 2: TyDiQA gold passage, 1-shot F1 lab | 35.8% ✎ | ▂··▁▃▂▂▂█▅▁▂ beats 3/25 | 1.5% |
$ where to run it
llama.cpp · recommended
version
- Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.macOS / Linux
curl -LsSf https://llama.app/install.sh | shWindowsirm https://llama.app/install.ps1 | iex - Run Tulu 2 DPO 70B · needs ~41.4 GB · from huggingface.co/TheBlokechat in the terminal
llama cli -hf TheBloke/tulu-2-dpo-70B-GGUF:Q4_K_M - Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf TheBloke/tulu-2-dpo-70B-GGUF:Q4_K_M
LM Studio
version
- Install LM Studio · Open the app once before using the lms command.
- Run Tulu 2 DPO 70B · needs ~41.4 GB · from huggingface.co/TheBloke
- Or download it in the terminal
lms get https://huggingface.co/TheBloke/tulu-2-dpo-70B-GGUF@Q4_K_M -y - Then run lms ls to see its name, and lms chat followed by that name
Ollama
version
- Install Ollama
download it (macOS / Windows) ↗
Linuxcurl -fsSL https://ollama.com/install.sh | sh - Run Tulu 2 DPO 70B · needs ~41.4 GB · from huggingface.co/TheBlokechat in the terminal
ollama run hf.co/TheBloke/tulu-2-dpo-70B-GGUF:Q4_K_M
Other ways to run it
- Run it as published · needs ~137.95 GB · from huggingface.co/allenai
uv run --with torch --with 'transformers<5' --with accelerate python -c "from transformers import pipeline; print(pipeline('text-generation', model='allenai/tulu-2-dpo-70b', torch_dtype='auto', device_map='auto')('Once upon a time', max_new_tokens=40)[0]['generated_text'])"
Weights and documentation: huggingface.co/allenai ↗
$ how the score was worked out · Tulu 2 — base and instruction… 30%, AlpacaEval 1.0 and 2.0 leader… 26%, Arena-Hard-Auto v0.1 leaderbo… 16% and 4 more
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 |
|---|---|---|
 Tulu 2 — base and instruction-tuned evaluations · lab-reported | 5 | 30% |
 AlpacaEval 1.0 and 2.0 leaderboards (tatsu-lab/alpaca_eval repository) | 2 | 26% |
| Arena-Hard-Auto v0.1 leaderboard (repository README, pre-v2) | 1 | 16% |
| WildBench v2 leaderboard data (Ai2) · lab-reported | 1 | 9% |
| Polyrating: A Cost-Effective and Bias-Aware Rating System for LLM Evaluation (arXiv:2409.00696v3) | 1 | 8% |
 LMArena leaderboard dataset | 1 | 8% |
 Open Portuguese LLM Leaderboard (archived) | 1 | 2% |
| prior · pull toward the middle | 0% | |
| weighted average | = 42.3 |
$ 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