Code Llama 34B Instruct
Llama · Llama 2 Community License
Meta's dedicated 34B coding model. Still competitive for code generation but being surpassed by newer models like Qwen 2.5 Coder 32B. Shorter context window (16K) is a limitation for large codebases.
- Parameters
- 33.7B
- Architecture
- Dense
- Context
- 16,384 tokens
- Released
- 2023-08-24
- Engines
- llama.cpp, ollama, vLLM
- Builder Tools
- Continue, Aider
Parameters
33.7B
VRAM
22.7 GB
Context
16K
Formats
4
GPUs
18
Code Llama 34B Instruct (33.7B) requires 22.7 GB VRAM at recommended quality (Q5_K_M). At efficient quality (Q4_K_M), it fits in 19 GB VRAM, making it compatible with the AMD Radeon RX 9060 XT 16GB. On NVIDIA Grace Blackwell Ultra GB300, expect approximately 135 tok/s at Q5_K_M. For the best experience, AMD AI Powerhouse ($1,818) is recommended.
Source: OwnRig methodology
22.7 GB
Q5_K_M
20.2 GB
16K tokens
Coding
VRAM Requirements
| Quality | Quantization | VRAM | File Size |
|---|---|---|---|
| recommended | Q5_K_M | 22.7 GB | 20.2 GB |
| efficient | Q4_K_M | 19 GB | 16.9 GB |
| compressed | Q3_K_M | 15.3 GB | 13.1 GB |
| compressed | Q2_K | 12 GB | 10.1 GB |
Context Length Impact
KV cache VRAM at Q5_K_M quality. Longer context = more memory.
| Context | KV Cache | Total VRAM |
|---|---|---|
| 2K | 410 MB | 23.1 GB |
| 4K | 819 MB | 23.5 GB |
| 8K | 1.5 GB | 24.2 GBexceeds 24 GB |
| 16K | 3.1 GB | 25.8 GBexceeds 24 GB |
Compatible GPUs
18 devices| NVIDIA Grace Blackwell Ultra GB300 | Q5_K_M | 135 tok/s | Excellent |
| Apple M4 Max (36GB Unified) | Q4_K_M | 14 tok/s | Good |
| NVIDIA GeForce RTX 4090 | Q4_K_M | 22 tok/s | Good |
| AMD Radeon RX 7900 XTX | Q4_K_M | 19 tok/s | Good |
| AMD Radeon Pro W7900 | Q4_K_M | 24 tok/s | Good |
| NVIDIA RTX PRO 6000 Blackwell | Q5_K_M | 48 tok/s | Good |
| NVIDIA RTX PRO 6000 Blackwell Max-Q | Q5_K_M | 44 tok/s | Good |
| AMD Radeon RX 9070 | Q2_K | – | Acceptable |
| AMD Radeon RX 9060 XT 16GB | Q2_K | – | Acceptable |
| AMD Radeon RX 7600 | Q2_K | 2 tok/s | Marginal |
| Apple M3 Pro (18GB Unified) | Q3_K_M | – | Not viable |
| NVIDIA GeForce RTX 3080 10GB | Q2_K | – | Not viable |
| NVIDIA GeForce RTX 4060 8GB | Q2_K | – | Not viable |
| NVIDIA RTX 4060 Laptop (40-60W) | Q2_K | – | Not viable |
| NVIDIA RTX 4070 Laptop (80-115W) | Q2_K | – | Not viable |
| NVIDIA GeForce RTX 4070 Ti 12GB | Q2_K | – | Not viable |
| AMD Radeon RX 9060 XT 8GB | Q2_K | – | Not viable |
| NVIDIA GeForce RTX 5060 8GB | Q2_K | – | Not viable |
Showing 18 of 18 entries
Builder Context
Code Llama 34B Instruct is commonly used with Continue, Aider. For an AI coding workflow, pair it with an embedding model like nomic-embed-text for local RAG.
Recommended Builds
Complete PC builds that can run Code Llama 34B Instruct.
Frequently Asked Questions
- How much VRAM does Code Llama 34B Instruct need?
- Code Llama 34B Instruct requires 22.7 GB VRAM at recommended quality (Q5_K_M). At lower quality settings, it can fit in as little as 12 GB.
- What is the best GPU for Code Llama 34B Instruct?
- The NVIDIA Grace Blackwell Ultra GB300 delivers the best performance for Code Llama 34B Instruct, achieving 135 tok/s at Q5_K_M with an excellent rating.
- What quantization should I use for Code Llama 34B Instruct?
- For the best quality, use Q5_K_M (22.7 GB VRAM). If your GPU has limited VRAM, Q2_K (12 GB) is the most efficient option with acceptable quality.
- Is Code Llama 34B Instruct good for coding?
- Yes. Code Llama 34B Instruct is used with Continue, Aider for local AI coding. For the best coding experience, pair it with an embedding model for local RAG.
Data confidence: verified. 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.Llama is a trademark of its respective owner. OwnRig is not affiliated with or endorsed by the model creator.