Llama 4 Maverick 17B Instruct FP8 vs Mistral Large 3 675B Instruct
Llama 4 Maverick 17B Instruct FP8 (2025) and Mistral Large 3 675B Instruct (2025) are compact production models from AI at Meta and MistralAI. Llama 4 Maverick 17B Instruct FP8 ships a 1m-token context window, while Mistral Large 3 675B Instruct ships a 128k-token context window. On MMLU PRO, Mistral Large 3 675B Instruct leads by 5 pts. This comparison covers specs, pricing, API access, capabilities, benchmarks, input and output token costs, and production fit for coding and agent workloads.
Llama 4 Maverick 17B Instruct FP8 is ~233% cheaper at $0.15/1M; pay for Mistral Large 3 675B Instruct only for vision-heavy evaluation.
Decision scorecard
Local evidence first| Signal | Llama 4 Maverick 17B Instruct FP8 | Mistral Large 3 675B Instruct |
|---|---|---|
| Best for | multimodal apps, long-context analysis, and provider-routed production | multimodal apps and provider-routed production |
| Decision fit | Coding, RAG, and Agents | Coding, RAG, and Agents |
| Context window | 1m | 128k |
| Cheapest output | $0.60/1M tokens | $1.50/1M tokens |
| Provider routes | 10 tracked | 6 tracked |
| Shared benchmarks | 5 shared | MMLU PRO leader |
Decision tradeoffs
- Llama 4 Maverick 17B Instruct FP8 holds a shared-benchmark lead on Google-Proof Q&A, ahead by 23.2 points.
- Llama 4 Maverick 17B Instruct FP8 has the larger context window for long prompts, retrieval packs, or transcript analysis.
- Llama 4 Maverick 17B Instruct FP8 has the lower cheapest tracked output price at $0.60/1M tokens.
- Llama 4 Maverick 17B Instruct FP8 has broader tracked provider coverage for fallback and procurement flexibility.
- Local decision data tags Llama 4 Maverick 17B Instruct FP8 for Coding, RAG, and Agents.
- Mistral Large 3 675B Instruct holds a shared-benchmark lead on MMLU PRO, ahead by 5 points.
- Local decision data tags Mistral Large 3 675B Instruct for Coding, RAG, and Agents.
Monthly cost at traffic
Estimate token spend from the cheapest tracked input and output route or tier on this page.
Llama 4 Maverick 17B Instruct FP8
$270
Cheapest tracked route/tier: OpenRouter
Mistral Large 3 675B Instruct
$775
Cheapest tracked route/tier: OpenRouter
Estimated monthly gap: $505. Batch, cache, alternate speed tiers, and negotiated pricing are excluded from this local estimate.
Switch friction
- Provider overlap exists on OpenRouter, AWS Bedrock, and NVIDIA NIM; start route-level A/B tests there.
- Mistral Large 3 675B Instruct is $0.90/1M tokens higher on cheapest tracked output pricing, so quality gains need to justify the spend.
- Provider overlap exists on Microsoft Foundry, OpenRouter, and NVIDIA NIM; start route-level A/B tests there.
- Llama 4 Maverick 17B Instruct FP8 is $0.90/1M tokens lower on cheapest tracked output pricing before cache, batch, or negotiated discounts.
Specs
| Specification | ||
|---|---|---|
| Released | 2025-04-05 | 2025-12-01 |
| Context window | 1m | 128k |
| Parameters | 400B (17B active) | 675B |
| Architecture | Mixture of Experts | Decoder Only |
| License | Llama 4 Community | Apache 2.0OSI-approved |
| Openness | Open weights | Open source |
| Commercial use | Commercial use: conditional | Commercial use: permitted |
| Knowledge cutoff | 2024-08 | 2024-11 |
Pricing and availability
| Pricing attribute | Llama 4 Maverick 17B Instruct FP8 | Mistral Large 3 675B Instruct |
|---|---|---|
| Input price | $0.15/1M tokens | $0.50/1M tokens |
| Output price | $0.60/1M tokens | $1.50/1M tokens |
| Providers |
Capabilities
| Capability | Llama 4 Maverick 17B Instruct FP8 | Mistral Large 3 675B Instruct |
|---|---|---|
| Vision | Yes | Yes |
| Multimodal | Yes | Yes |
| Reasoning | No | No |
| Function calling | No | No |
| Tool use | No | No |
| Structured outputs | Yes | Yes |
| Code execution | No | No |
| IDE integration | No | No |
| Computer use | No | No |
| Parallel agents | No | No |
Benchmarks
| Benchmark | Llama 4 Maverick 17B Instruct FP8 | Mistral Large 3 675B Instruct |
|---|---|---|
| MMLU PRO | 80.5 | 85.5 |
| Google-Proof Q&A | 67.1 | 43.9 |
| LiveCodeBench | 43.4 | 82.8 |
| HumanEval | 77.4 | 92.0 |
| τ-bench | 68.5 | 70.2 |
Deep dive
On shared benchmark coverage, MMLU PRO has Llama 4 Maverick 17B Instruct FP8 at 80.5 and Mistral Large 3 675B Instruct at 85.5, with Mistral Large 3 675B Instruct ahead by 5 points; Google-Proof Q&A has Llama 4 Maverick 17B Instruct FP8 at 67.1 and Mistral Large 3 675B Instruct at 43.9, with Llama 4 Maverick 17B Instruct FP8 ahead by 23.2 points; LiveCodeBench has Llama 4 Maverick 17B Instruct FP8 at 43.4 and Mistral Large 3 675B Instruct at 82.8, with Mistral Large 3 675B Instruct ahead by 39.4 points. The largest visible gap is 39.4 points on LiveCodeBench, which matters most when that benchmark mirrors your workload. Treat isolated benchmark wins as directional, because provider routing, prompt style, and tool access can move real application results.
The capability footprint is close: both models cover vision, multimodal input, and structured outputs. That makes context budget, benchmark fit, and provider maturity more important than a simple checklist. If your application depends on one integration detail, verify it against the provider route you plan to use, not just the base model listing.
For cost, Llama 4 Maverick 17B Instruct FP8 lists $0.15/1M input and $0.60/1M output tokens on the cheapest tracked provider, while Mistral Large 3 675B Instruct lists $0.50/1M input and $1.50/1M output tokens on the cheapest tracked provider. A 70/30 input-output blend puts Llama 4 Maverick 17B Instruct FP8 lower by about $0.51 per million blended tokens. Availability is 10 providers versus 6, so concentration risk also matters.
Choose Llama 4 Maverick 17B Instruct FP8 when long-context analysis, larger context windows, and lower input-token cost are central to the workload. Choose Mistral Large 3 675B Instruct when vision-heavy evaluation are more important. For production, rerun your own prompts through the exact provider, region, and tool stack you plan to ship.
FAQ
Which has a larger context window, Llama 4 Maverick 17B Instruct FP8 or Mistral Large 3 675B Instruct?
Llama 4 Maverick 17B Instruct FP8 supports 1m tokens, while Mistral Large 3 675B Instruct supports 128k tokens. That gap matters most for long documents, large codebases, retrieval-heavy agents, and conversations where earlier context must remain visible.
Which is cheaper, Llama 4 Maverick 17B Instruct FP8 or Mistral Large 3 675B Instruct?
Llama 4 Maverick 17B Instruct FP8 is cheaper on tracked token pricing. Llama 4 Maverick 17B Instruct FP8 costs $0.15/1M input and $0.60/1M output tokens. Mistral Large 3 675B Instruct costs $0.50/1M input and $1.50/1M output tokens. Provider discounts or batch pricing can still change the final bill.
Is Llama 4 Maverick 17B Instruct FP8 or Mistral Large 3 675B Instruct open source?
Llama 4 Maverick 17B Instruct FP8 is listed under Llama 4 Community. Mistral Large 3 675B Instruct is listed under Apache 2.0. License labels affect whether you can self-host, redistribute weights, or rely only on hosted APIs, so confirm the upstream license before deployment.
Which is better for vision, Llama 4 Maverick 17B Instruct FP8 or Mistral Large 3 675B Instruct?
Both Llama 4 Maverick 17B Instruct FP8 and Mistral Large 3 675B Instruct expose vision. The better choice depends on benchmark fit, context budget, pricing, and whether your provider route exposes the same capability surface.
Which is better for multimodal input, Llama 4 Maverick 17B Instruct FP8 or Mistral Large 3 675B Instruct?
Both Llama 4 Maverick 17B Instruct FP8 and Mistral Large 3 675B Instruct expose multimodal input. The better choice depends on benchmark fit, context budget, pricing, and whether your provider route exposes the same capability surface.
Where can I run Llama 4 Maverick 17B Instruct FP8 and Mistral Large 3 675B Instruct?
Llama 4 Maverick 17B Instruct FP8 is available on Microsoft Foundry, Together AI, OpenRouter, Fireworks AI, and DeepInfra. Mistral Large 3 675B Instruct is available on OpenRouter, AWS Bedrock, NVIDIA NIM, Mistral AI Studio, and Microsoft Foundry. Provider coverage can affect latency, region availability, compliance posture, and fallback options.
Continue comparing
Last reviewed: 2026-06-19. Data sourced from public model cards and provider documentation.