Alternatives · 15 models compared · checked 17 Aug 2026
Mistral Small 3 alternatives: 15 models in or near grid B2
The closest model to Mistral Small 3 on the atlas is Llama 3.2 1B Instruct, in the same square, at $0.03 per 1M input tokens against Mistral Small 3’s . 5 of the 15 nearest models are cheaper on input; the cheapest is gpt-oss-20b at $0.02. We rank by position on the atlas, not by capability, because we hold no capability data.
$0.05
observed We measured
- Source
- Our catalogue
- Retrieved
- 17 Aug 2026
Ask an AI about this pageGChatGPTCClaudePPerplexityGeGeminiXGrok
Cite
Cite this page
ModelsAtlas. “Mistral Small 3 alternatives: 15 models in or near grid B2”. Retrieved 17 Aug 2026. https://modelsatlas.com/models/mistralai/mistral-small-24b-instruct/alternatives
Where the alternatives sit
Each square is a price band (A–H) × a context band (1–6). How the grid works
Same squareGRIDB22 modelsCommand R7B (12-2024) and Llama 3.2 1B Instruct
Mistral Small 3Its 15 nearest modelsOther modelsOpen weights
View the map as a table (4 squares)
| Square | Price band | Context band | Models |
|---|---|---|---|
| GRIDB1 | <$0.10 | <32K | Phi 4 |
| GRIDB2 | <$0.10 | 32–128K | Mistral Small 3, Llama 3.2 1B Instruct, Command R7B (12-2024) |
| GRIDB3 | <$0.10 | 128–256K | gpt-oss-20b, Mistral Nemo, gpt-oss-120b, Gemma 3 12B, Gemma 3 4B, Llama 3.1 8B Instruct, Llama 3.2 3B Instruct, gpt-oss-safeguard-20b, Gemma 3 27B, Qwen3 32B, Mistral Small 3.2 24B |
| GRIDC2 | $0.10–0.30 | 32–128K | Qwen2.5 7B Instruct |
Top 15 alternatives
Ranked by steps away on the atlas, then cheaper before pricier, then input price
| # | Model | Grid | Why it fits | Input / 1M | vs Mistral Small 3 | Output / 1M | Context | |
|---|---|---|---|---|---|---|---|---|
| — | Mistral Small 3Mistral AI · the model you’re replacing | GRIDB2 | Baseline | $0.05 | — | $0.08 | 33K | |
| 1 | Llama 3.2 1B InstructMeta | GRIDB2 | Same square, 46% cheaper input, 1.8× the context | $0.03 | −46% | $0.20 | 60K | Compare → |
| 2 | Command R7B (12-2024)Cohere | GRIDB2 | Same square, 25% cheaper input, 3.9× the context | $0.04 | −25% | $0.15 | 128K | Compare → |
| 3 | gpt-oss-20bOpenAI | GRIDB3 | Neighbouring square B3, 64% cheaper input, 4× the context | $0.02 | −64% | $0.09 | 131K | |
| 4 | Mistral NemoMistral AI | GRIDB3 | Same maker, neighbouring square B3, 42% cheaper input, 4× the context | $0.03 | −42% | $0.03 | 131K | |
| 5 | gpt-oss-120bOpenAI | GRIDB3 | Neighbouring square B3, 40% cheaper input, 4× the context | $0.03 | −40% | $0.15 | 131K | |
| 6 | Gemma 3 12BGoogle | GRIDB3 | Neighbouring square B3, same input price, 4× the context | $0.05 | same | $0.15 | 131K | Compare → |
| 7 | Gemma 3 4BGoogle | GRIDB3 | Neighbouring square B3, same input price, 4× the context | $0.05 | same | $0.10 | 131K | Compare → |
| 8 | Llama 3.1 8B InstructMeta | GRIDB3 | Neighbouring square B3, same input price, 4× the context | $0.05 | same | $0.08 | 131K | Compare → |
| 9 | Llama 3.2 3B InstructMeta | GRIDB3 | Neighbouring square B3, same input price, 4× the context | $0.05 | same | $0.33 | 131K | Compare → |
| 10 | Phi 4Microsoft | GRIDB1 | Neighbouring square B1, 40% pricier input | $0.07 | +40% | $0.14 | 16K | Compare → |
| 11 | gpt-oss-safeguard-20bOpenAI | GRIDB3 | Neighbouring square B3, 50% pricier input, 4× the context | $0.08 | +50% | $0.30 | 131K | |
| 12 | Gemma 3 27BGoogle | GRIDB3 | Neighbouring square B3, 60% pricier input, 4× the context | $0.08 | +60% | $0.45 | 131K | |
| 13 | Qwen3 32BQwen (Alibaba) | GRIDB3 | Neighbouring square B3, 60% pricier input, 4× the context | $0.08 | +60% | $0.28 | 131K | |
| 14 | Mistral Small 3.2 24BMistral AI | GRIDB3 | Same maker, neighbouring square B3, 88% pricier input, 7.8× the context | $0.09 | +88% | $0.25 | 256K | |
| 15 | Qwen2.5 7B InstructQwen (Alibaba) · open weights | GRIDC2 | Neighbouring square C2, 100% pricier input, open weights | $0.10 | +100% | $0.20 | 33K |
What switching saves
3,000 input + 800 output tokens × 1,000 requests a day
Σ Derived from standard-tier prices; excludes caching and batch discounts. Use your own numbers →
Larger-context alternatives
| Model | Input / 1M | Context | |
|---|---|---|---|
| Llama 3.2 1B InstructMeta | $0.03 | 60K | Compare |
| Command R7B (12-2024)Cohere | $0.04 | 128K | Compare |
| gpt-oss-20bOpenAI | $0.02 | 131K | |
| Mistral NemoMistral AI | $0.03 | 131K | |
| gpt-oss-120bOpenAI | $0.03 | 131K |
Free alternatives
| Model | Input / 1M | Context | |
|---|---|---|---|
| Lyria 3 Clip PreviewGoogle | Free | 1.05M | |
| Lyria 3 Pro PreviewGoogle | Free | 1.05M |
Compare Mistral Small 3 head to head
Questions
What is the best alternative to Mistral Small 3?
Llama 3.2 1B Instruct ranks first: same square, 46% cheaper input, 1.8× the context. Input is $0.03 per 1M tokens against $0.05 for Mistral Small 3. Ranking is by position on the atlas, not by quality.
What is the cheapest alternative to Mistral Small 3?
gpt-oss-20b at $0.02 per 1M input tokens, 64% less than Mistral Small 3, among the 15 closest.
Is there an open-weight alternative to Mistral Small 3?
Yes, 1 of the 15: Qwen2.5 7B Instruct.
Which alternative has a larger context window?
Mistral Small 3.2 24B, gpt-oss-20b, Mistral Nemo, gpt-oss-120b and 9 more hold more than Mistral Small 3's 33K; the largest is Mistral Small 3.2 24B at 256K.
How are alternatives ranked?
By distance on the atlas grid from Mistral Small 3's square B2: the same square first, then neighbouring squares. Within the same distance, cheaper models come before pricier ones, then the lowest input price wins. We hold no capability data, so we do not claim a model can do what Mistral Small 3 does, only that it costs and holds about the same. No vendor pays for position.
More alternatives
Keep going in your AI
Opens with this page as the source. We add nothing else to the prompt.
- What is the best alternative to Mistral Small 3?
- What is the cheapest alternative to Mistral Small 3?
- Is there an open-weight alternative to Mistral Small 3?