Olmo 3.1 32B Think
Olmo 3.1 32B Think is Ai2's fully open reasoning model, updated in December 2025, with weights, training data, code and intermediate checkpoints all published. It is a dense 32B model that writes long chain-of-thought for math and coding. Ai2 releases it under Apache 2.0 for research and education.
Specifications
- Context window
- 64Kabout 49,200 words, or 98 pages
- Max output
- Not published
- Input price
- —no list price confirmed
- Output price
- —
- Cached input
- Not published
- Weights
- OpenApache-2.0
- Parameters
- 32Bdense
- Input types
- Text
- Output types
- Text
- Reasoning mode
- Yes
- Tool calling
- Not documented
- Released
- Dec 12, 2025Current
No serverless price found on the official pricing pages of Together AI, Fireworks, DeepInfra or Groq as of 2026-10-10; Ai2 lists OpenRouter and its own playground as access points.
What it costs in practice
| Workload | Monthly cost at list price |
|---|---|
| Support chatbot30,000 conversations a month, 2,000 input and 400 output tokens each | No list price |
| Document summaries2,000 documents a month, 12,000 input and 800 output tokens each | No list price |
| Coding agent1,000 tasks a month, 60,000 input and 6,000 output tokens each | No list price |
Standard rates, no caching or batch discounts. Run your own numbers.
Strengths
- Training data (Dolma 3, Dolci), code and checkpoints all released
- Apache 2.0 license
- Well suited to reinforcement-learning research
Trade-offs to weigh
- 65K context window
- English-focused and text-only; tool use is not documented for this variant
Where it fits
Ai2 (Allen Institute for AI) positions Olmo 3.1 32B Think for:
Where you can use it
- Hugging Face
- Ollama
- OpenRouter
- Ai2 Playground
Run it yourself
Released under Apache-2.0. Approximate memory needed just to load the weights:
| Precision | Memory for weights |
|---|---|
| 16-bit (BF16/FP16) | 64 GB |
| 8-bit (FP8/INT8) | 32 GB |
| 4-bit (Q4_K_M, about 4.85 bits) | 19 GB |
Add memory for the KV cache (it grows with context length and concurrent users) and 1-2 GB of runtime overhead. Size it precisely in the VRAM calculator.
Alternatives to consider
| Model | Tier | Context | Input / output per 1M | |
|---|---|---|---|---|
| Muse Glimmer 30BMeta | Efficient | 128K | $0.35 / $1.50 | Compare |
| Mistral Small 4Mistral AI | Efficient | 256K | $0.15 / $0.60 | Compare |
| Ministral 3 14BMistral AI | Efficient | 256K | $0.20 / $0.20 | Compare |
| Qwen3.8-27BAlibaba (Qwen) | Efficient | 256K | $0.80 / $4 | Compare |
Frequently asked questions
How much does Olmo 3.1 32B Think cost?
We couldn't confirm a published per-token list price for Olmo 3.1 32B Think on October 10, 2026. No serverless price found on the official pricing pages of Together AI, Fireworks, DeepInfra or Groq as of 2026-10-10; Ai2 lists OpenRouter and its own playground as access points. Check the provider's pricing page for current rates.
What is the context window of Olmo 3.1 32B Think?
Olmo 3.1 32B Think accepts up to 65,536 tokens of context (about 49,200 words, or 98 pages). Long prompts cost more and can be slower, so most apps send only what the task needs.
Is Olmo 3.1 32B Think open source?
Olmo 3.1 32B Think is an open-weight model released under the Apache-2.0 license. You can download the weights from Hugging Face (allenai/Olmo-3.1-32B-Think) and run or fine-tune them yourself, within the license terms. Open weights isn't the same as fully open source: training data and code may not be published.
Can I run Olmo 3.1 32B Think locally?
Yes, if you have enough memory. The weights alone need about 19 GB at 4-bit quantization and about 64 GB at 16-bit, plus room for the KV cache. Use the VRAM calculator on LLModel.com to check a specific GPU or Mac.
What is Olmo 3.1 32B Think best for?
Ai2 (Allen Institute for AI) positions Olmo 3.1 32B Think for research, reasoning, fine-tuning, self-hosting. It accepts text as input, has a reasoning (thinking) mode and doesn't document tool calling. Test it on a sample of your own tasks before you choose it for production.
Sources
- huggingface.co/allenai/Olmo-3.1-32B-Think
- allenai.org/blog/olmo3
- allenai.org/olmo
- huggingface.co/allenai/Olmo-3.1-32B-Think/raw/main/config.json
- ollama.com/library
Facts checked against these pages on October 10, 2026. Spot an error? Tell us.
Still choosing a model?
Send us your use case and constraints. We'll come back with a shortlist and an evaluation plan you can run on your own data.
- Free and vendor-neutral: no model provider pays for placement.
- Your details go only to partners you agree to.