Mistral Nemo
Mistral AI · released Jul 17, 2024 · mistralai/Mistral-Nemo-Instruct-2407
- Type
- Open weightsApache License 2.0
- Params
- 12.2B
- Context
- 131K
about 98K words of context
Our take
Written Sep 2, 2026Mistral Nemo is a 12.2-billion-parameter text model released in 2024 with a permissive Apache licence. It offers a 131,072-token request limit and some of the cheapest hosted inference among open-weights models we track, though its reasoning scores are near the floor for measured models.
Pick this for budget text generation where licence permissiveness matters, or for long-context tasks up to 131,072 tokens on a downloadable model. Use it when instruction-following accuracy in the low-sixties is sufficient. Skip it if you need graduate-level science reasoning, strong broad knowledge, or guaranteed full usability of the full context length.
The case for it
- Extremely cheap hosted inference for a downloadable model: third-party hosts charge a small fraction of the vendor's own rate with only modest throughput reduction.
- Apache 2.0 licence allows commercial use, fine-tuning and redistribution without restriction.
- 131,072-token request limit is large for a 12.2-billion-parameter model.
The case against it
- Near-floor performance on graduate-level science questions, at 5.4% correct on GPQA Diamond.
- Weak broad knowledge: 28% correct on the harder multi-task benchmark.
- Vendor hosting is drastically overpriced versus identical third-party offers for the same weights.
How good is it?
EverydayGeneral questions and everyday reasoning
Not yet scored on Arena Text (overall). It is on GPQA Diamond, in 12th of 16 with 29.
CodingWriting and fixing code on its own
Not yet scored on Arena Coding.
AgenticPlanning, calling tools, staying on task
Not yet scored on Arena Agent.
WritingDrafting and rewriting prose
Not yet scored on Arena Creative Writing.
These tests check whether a model follows instructions — a precondition for all the work above, but not a measure of how well that work is done.
Every published score for this model3 scoresEvery figure we hold, from 3 boards, with who ran it and a link to the source — including the boards no rating above is built on.
Can you run it yourself?
Comfortable fit
GeForce RTX 4090 · 24 GB
Room to spare. 13.3 GB spare means a 10% error in the size would not change the answer.
GeForce RTX 5090 · 32 GB
Room to spare. 21.3 GB spare means a 10% error in the size would not change the answer.
Apple M1 (8-core GPU) · 16 GB
Room to spare. 2.5 GB spare means a 10% error in the size would not change the answer.
Memory use by level
Against a 24 GB card.
What is quantisation? →This model on every device we track71 devicesThe Q4 build most people download, on each device: what the weights come to, how much context the memory leaves, and whether it runs. Smallest device that runs it first.
Check against your own machine → · Where to rent it hosted →
Or rent it from someone else
Prices checked between 4 hours and 10 hours ago — each listing carries its own date.
- per 1M tokens
- $0.019 in / $0.030 out
- Context served
- 131K
- Throughput
- Not measured
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| OpenRouterOpenRouter's own listing | $0.019 / $0.030checked 4 hours ago | 131K | not measured | Unknown | Unknown | Unknown |
| DeepInfrafp8Direct and through OpenRouter | $0.019 / $0.030checked 4 hours ago | 131K16K max reply through OpenRouter | 19 tok/sthrough OpenRouter | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterConfirmed |
| Parasailfp8Through OpenRouter | $0.030 / $0.030checked 4 hours ago | 131K105K max reply | 36 tok/s | No | No | Confirmed |
| DekaLLMfp8Through OpenRouter | $0.018 / $0.030checked 4 hours ago | 131K105K max reply | 13 tok/s | No | No | Confirmed |
| Io Netfp16Through OpenRouter | $0.036 / $0.13checked 4 hours ago | 128K102K max reply | 19 tok/s | No | No | Confirmed |
| Mistral AIeuThrough OpenRouter | $0.15 / $0.15checked 4 hours ago | 131K105K max reply | 9 tok/s | No | Yes30 days | Confirmed |
| Novita AIfp8Direct and through OpenRouter | $0.040 / $0.17checked 4 hours ago directchecked 10 hours ago through OpenRouter | 60K16K max reply through OpenRouter | 59 tok/sthrough OpenRouter | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterConfirmed |
DeepInfra
Mistral AI
Novita AIAcross the 7 listings we hold: 6 say they do not train on prompts (2 of them only through OpenRouter), 0 say they do and 1 does not say. 6 appear in the zero-retention registry we check (2 of them only through OpenRouter); the rest are unknown to us.
What each host's API supports
From the parameter list each endpoint publishes. Streaming is omitted: nothing we hold reports it, for any model.
| Provider | Tool calling | JSON output | Strict schema |
|---|---|---|---|
| OpenRouterOpenRouter's own listing | ✓ | ✓ | ✓ |
| DeepInfrafp8Direct and through OpenRouter | ✓ | ✓ | ✓ |
| Parasailfp8Through OpenRouter | ✗ | ✓ | ✓ |
| DekaLLMfp8Through OpenRouter | ✗ | ✓ | ✓ |
| Io Netfp16Through OpenRouter | ✓ | ✗ | ✗ |
| Mistral AIeuThrough OpenRouter | ✓ | ✓ | ✓ |
| Novita AIfp8Direct and through OpenRouter | ✗ | ✓ | ✓ |
Tool calling: 4 of 7 listings say yes, 3 say no. JSON output: 6 of 7 listings say yes, 1 says no. Strict schema: 6 of 7 listings say yes, 1 says no.
Models people weigh against Mistral Nemo
When we formed this view
Recent changes
What moved
input −8% ($0.0440 → $0.0405 per 1M tokens), output −8% ($0.160 → $0.147 per 1M tokens), cache read −8% ($0.0290 → $0.0267 per 1M tokens)What moved
input +5% ($0.042 → $0.044 per 1M tokens), cache read +32% ($0.022 → $0.029 per 1M tokens)What moved
input −2% ($0.043 → $0.042 per 1M tokens), output −3% ($0.165 → $0.160 per 1M tokens), cache read +3% ($0.02145 → $0.02200 per 1M tokens)What moved
input +10% ($0.039 → $0.043 per 1M tokens), output +10% ($0.150 → $0.165 per 1M tokens), cache read +10% ($0.0195 → $0.0215 per 1M tokens)What moved
input +11% ($0.035 → $0.039 per 1M tokens), output +15% ($0.13 → $0.15 per 1M tokens), cache read +11% ($0.0175 → $0.0195 per 1M tokens)What moved
first indexed by our pipelineEach date is the day we first saw the change, or the day the maker announced it.
What we do not know about this model yet
- We hold no measured file for it, so all 3 sizes on this page are calculated from the parameter count.
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- 1 of 7 listings does not say whether it trains on prompts, and 2 answer only through OpenRouter, not for their own listing.
- We hold no batch or off-peak rate for any of its listings.
- We hold a decode speed for it, but no prompt-processing (prefill) figure, so how long the input side of a job takes is unknown to us.
Licence and identifiers
What the licence allowsApache License 2.0, what it allows commercially, and the identifiers you need to pull this model — its Hugging Face repo, our slug and a machine-readable card.
Licence
Apache License 2.0
Fully permissive: commercial use, redistribution, and derivatives allowed. Requires attribution and a copy of the license. Includes an express patent grant.
Identifiers
- Hugging Face
- mistralai/Mistral-Nemo-Instruct-2407
- Architecture
- Dense
- Takes in, gives back
- Text in, text out
- Catalogue slug
- mistralai-mistral-nemo