Nanbeige 4.3 33B
Nanbeige's flagship 33B dense language and coding model with a 512K context window. Engineered specifically for agentic terminal workflows, SWE-bench engineering, and professional enterprise co-work under Apache 2.0.
The brutal truth
A serious alternative to Qwen2.5-Coder-32B built specifically for autonomous agents that run terminal commands and refactor full repos. If you run SWE-bench agents on a 24GB card, this is a top-tier driver.
Files and Versions
Direct inspected tree artifacts from Hugging Face. Verified filenames, exact byte sizes, model weight shards, and configuration definitions.
Decision matrix
✓ RUN / USE THIS IF…
- •You build autonomous developer agents using Claude Code, Cline, or Aider
- •You own a 24GB GPU (RTX 3090/4090 or Mac Studio)
- •Your 24GB GPU comfortably fits the model weights (~21GB VRAM) with dedicated headroom for inference.
- •You require 100% offline data privacy with zero third-party API logging, subscriptions, or prompt retention.
✕ WALK AWAY IF…
- •You have less than 24GB VRAM (use Spark-X2.6 or Qwen-14B instead)
- •You only need casual conversational chat
- •Your GPU has less than 21GB VRAM and you cannot tolerate slow CPU/system RAM offload paging.
- •You prefer a plug-and-play web app interface and don't want to manage local model checkpoints, drivers, or runtimes.
True cost of ownership
About
Nanbeige's flagship 33B dense language and coding model with a 512K context window. Engineered specifically for agentic terminal workflows, SWE-bench engineering, and professional enterprise co-work under Apache 2.0.
What it does
Download the weights from Hugging Face, pick a quantisation that matches your GPU, and run it locally with Ollama, llama.cpp, vLLM, or a compatible runtime. Agentic terminal workflows & 512K context coding runs on your own hardware — no per-token bill, no third-party data handling.
Key features
Excels at agentic coding and bash terminal execution on Terminal-Bench and DeepSWE
Vast 512K token context window for reading massive codebases and enterprise repos
Fully open Apache 2.0 license with configurable thinking/reasoning modes
Use cases
Inline edits, refactors, and test generation inside the editor.
Flag null-safety, security, and style issues before merge.
Turn long procedural functions into typed, testable modules.
Generate integration tests from route definitions and schemas.
Quantization & Hardware Fit
VRAM Hardware Compatibility & Memory Allocation
| Quant | VRAM | Quality / Fit on 24GB | Install / Run Command (OLLAMA) | |
|---|---|---|---|---|
| Q4_K_M● ACTIVE | 21 GB | Standard · ✓ Fits (3GB spare) | ollama run nanbeige-43-33b | |
| Q5_K_M | 24 GB | Standard · ✓ Fits (0GB spare) | ollama run nanbeige-43-33b | |
| Q8_0 | 36 GB | Standard · ✕ Needs 36GB | ollama run nanbeige-43-33b | |
| FP16 | 66 GB | Standard · ✕ Needs 66GB | ollama run nanbeige-43-33b |
Buyer doubts, cleared
Can I actually run Nanbeige 4.3 33B on my 24GB GPU?
What's the fine print on the license?
Does the 512K tokens actually hold up in practice?
User reviews
We evaluated Nanbeige 4.3 33B against two hosted alternatives. It won on cost-per-output but lost on latency. For batch work it is a clear yes.
The output reads naturally once you tune the prompt. It has cut my first-draft time roughly in half.
Clean interface, consistent results once you land on a prompt you like. Wish there were more export presets for client handoffs.
Switched our internal workflow to Nanbeige 4.3 33B after the last pricing change. Nobody noticed a quality difference, and the bill dropped by half.
