Mistral Large 2 123B
Mistral · Apache 2.0
Mistral's flagship 123B parameter model. Wide multilingual performance and code generation. 128K context window. Requires multi-GPU or high-memory Apple Silicon for home use. At Q3/Q2, fits on 2x RTX 4090 or 64 GB+ Apple Silicon.
- Parameters
- 123B
- Architecture
- Dense
- Context
- 128,000 tokens
- Released
- 2024-07-24
- Engines
- llama.cpp, vLLM, ollama
- Builder Tools
- Claude Code, Codex CLI, Continue, Cursor, LM Studio, Open WebUI, Windsurf
Parameters
123B
VRAM
95 GB
Context
125K
Formats
6
GPUs
14
Mistral Large 2 123B (123B) requires 95 GB VRAM at recommended quality (Q6_K). On NVIDIA Grace Blackwell Ultra GB300, expect approximately 30 tok/s at Q8_0. For the best experience, Mac Studio AI Builder ($3,999) is recommended.
Source: OwnRig methodology
95 GB
Q6_K
90 GB
125K tokens
Chat
VRAM Requirements
| Quality | Quantization | VRAM | File Size |
|---|---|---|---|
| full | Q8_0 | 130 GB | 123 GB |
| recommended | Q6_K | 95 GB | 90 GB |
| recommended | Q5_K_M | 82 GB | 77 GB |
| efficient | Q4_K_M | 70 GB | 65 GB |
| compressed | Q3_K_M | 56 GB | 52 GB |
| compressed | Q2_K | 45 GB | 42 GB |
Context Length Impact
KV cache VRAM at Q6_K quality. Longer context = more memory.
| Context | KV Cache | Total VRAM |
|---|---|---|
| 2K | 512 MB | 95.5 GBexceeds 24 GB |
| 4K | 1 GB | 96 GBexceeds 24 GB |
| 8K | 2 GB | 97 GBexceeds 24 GB |
| 16K | 4.1 GB | 99.1 GBexceeds 24 GB |
| 32K | 8.2 GB | 103.2 GBexceeds 24 GB |
| 64K | 16.4 GB | 111.4 GBexceeds 24 GB |
Compatible GPUs
14 devices| NVIDIA Grace Blackwell Ultra GB300 | Q8_0 | 30 tok/s | Good |
| Apple M4 Max (128GB Unified) | Q4_K_M | 10 tok/s | Acceptable |
| Apple M4 Ultra (192GB) | Q4_K_M | 15 tok/s | Acceptable |
| AMD Radeon Pro W7900 | Q2_K | 5 tok/s | Acceptable |
| NVIDIA RTX PRO 6000 Blackwell | Q5_K_M | 13 tok/s | Acceptable |
| NVIDIA RTX PRO 6000 Blackwell Max-Q | Q5_K_M | 12 tok/s | Acceptable |
| Apple M4 Max (64GB Unified) | Q2_K | 5 tok/s | Marginal |
| NVIDIA GeForce RTX 4090 | Q2_K | 3 tok/s | Marginal |
| NVIDIA GeForce RTX 5090 | Q3_K_M | 4 tok/s | Marginal |
| AMD Radeon RX 7600 | Q2_K | – | Not viable |
| AMD Radeon RX 7900 XTX | Q2_K | – | Not viable |
| AMD Radeon RX 9070 | Q2_K | – | Not viable |
| AMD Radeon RX 9060 XT 16GB | Q2_K | – | Not viable |
| AMD Radeon RX 9060 XT 8GB | Q2_K | – | Not viable |
Showing 14 of 14 entries
Builder Context
Mistral Large 2 123B is commonly used with Claude Code, Codex CLI, Continue, Cursor, LM Studio, Open WebUI, Windsurf. For an AI coding workflow, pair it with an embedding model like nomic-embed-text for local RAG.
Frequently Asked Questions
- How much VRAM does Mistral Large 2 123B need?
- Mistral Large 2 123B requires 95 GB VRAM at recommended quality (Q6_K). At lower quality settings, it can fit in as little as 45 GB.
- What is the best GPU for Mistral Large 2 123B?
- The NVIDIA Grace Blackwell Ultra GB300 delivers the best performance for Mistral Large 2 123B, achieving 30 tok/s at Q8_0 with an good rating.
- What quantization should I use for Mistral Large 2 123B?
- For the best quality, use Q6_K (95 GB VRAM). If your GPU has limited VRAM, Q2_K (45 GB) is the most efficient option with acceptable quality.
- Is Mistral Large 2 123B good for coding?
- Yes. Mistral Large 2 123B is used with Claude Code, Codex CLI, Continue, Cursor, LM Studio, Open WebUI, Windsurf for local AI coding. For the best coding experience, pair it with an embedding model for local RAG.
Data confidence: estimated. Source
VRAM requirements are calculated from model parameters and may vary by inference engine, context length, and batch size. Performance estimates are based on community benchmarks and should be verified for your specific configuration.Mistral is a trademark of its respective owner. OwnRig is not affiliated with or endorsed by the model creator.