November 21, 2024

Details matter with open source models

Open source models like Qwen 2.5 32B Instruct are performing very well on aider’s code editing benchmark, rivaling closed source frontier models.

But pay attention to how your model is being served and quantized, as it can impact code editing skill. Open source models are often available at a variety of quantizations, and can be served with different token limits. These details matter when working with code.

The graph above and table below compares different versions of the Qwen 2.5 Coder 32B Instruct model, served both locally and from a variety of cloud providers.

Pitfalls and details

This benchmarking effort highlighted a number of pitfalls and details specific to open source models which can have a significant impact on their ability to correctly edit code:

  • Quantization – Open source models are often available at dozens of different quantizations. Most seem to only modestly decrease code editing skill, but stronger quantizations do have a real impact.
  • Context window – Cloud providers can decide how large a context window to accept, and they often choose differently. Ollama’s local API server defaults to a tiny 2k context window, and silently discards data that exceeds it. Such a small window has catastrophic effects on performance, without throwing obvious hard errors.
  • Output token limits – Open source models are often served with wildly differing output token limits. This has a direct impact on how much code the model can write or edit in a response.
  • Buggy cloud providers – While benchmarking Qwen 2.5 Coder 32B Instruct and DeepSeek V2.5, I discovered multiple cloud providers with broken or buggy API endpoints. They seemed to be returning results different from expected based on the advertised quantization and context sizes. The harm caused to the code editing benchmark varied from serious to catastrophic. One provider scored 0.5% on the benchmark with DeepSeek V2.5, a highly capable model.

Closed source, proprietary models don’t typically have these issues. They are owned and operated by the organization that created them, and typically served with specific, predictable context window and output token limits. Their quantization level is usually unknown, but fixed and unchanging for all users.

Conclusions

The best versions of the Qwen model rival GPT-4o, while the worst performing quantization is more like the older GPT-4 Turbo when served competently. Even an otherwise excellent fp16 quantization falls to GPT-3.5 Turbo levels of performance if run with Ollama’s default 2k context window.

Sections

Benchmark results

These are results from single benchmark runs, so expect normal variance of +/- 1-2%.

Model Percent completed correctly Percent using correct edit format Command Edit format
Fireworks: unknown 72.2% 94.0% aider --model fireworks_ai/accounts/fireworks/models/qwen2p5-coder-32b-instruct diff
Deepinfra: BF16 72.2% 94.7% aider --model deepinfra/Qwen/Qwen2.5-Coder-32B-Instruct diff
mlx-community: 8bit 72.2% 92.5% aider --model openai/mlx-community/Qwen2.5-Coder-32B-Instruct-8bit diff
mlx-community: 4bit 72.2% 88.7% aider --model openai/mlx-community/Qwen2.5-Coder-32B-Instruct-4bit diff
Ollama: fp16 71.4% 90.2% aider --model ollama/qwen2.5-coder:32b-instruct-fp16 diff
HuggingFace via GLHF: BF16 71.4% 94.7% aider --model openai/hf:Qwen/Qwen2.5-Coder-32B-Instruct --openai-api-base https://glhf.chat/api/openai/v1 diff
Deepinfra via OpenRouter: BF16 69.9% 89.5% aider --model openrouter/qwen/qwen-2.5-coder-32b-instruct diff
Hyperbolic: BF16 69.2% 91.7% aider --model openai/Qwen/Qwen2.5-Coder-32B-Instruct --openai-api-base https://api.hyperbolic.xyz/v1/ diff
Hyperbolic via OpenRouter: BF16 68.4% 89.5% aider --model openrouter/qwen/qwen-2.5-coder-32b-instruct diff
Fireworks via OpenRouter: unknown 67.7% 94.0% aider --model openrouter/qwen/qwen-2.5-coder-32b-instruct diff
OpenRouter: multiple 67.7% 95.5% aider --model openrouter/qwen/qwen-2.5-coder-32b-instruct diff
Ollama: q4_K_M 66.9% 94.0% aider --model ollama/qwen2.5-coder:32b-instruct-q4_K_M diff
Ollama: q2_K 61.7% 91.7% aider --model ollama/qwen2.5-coder:32b-instruct-q2_K diff
Ollama: fp16, 2k ctx 51.9% 46.2% aider --model ollama/qwen2.5-coder:32b-instruct-fp16 # num_ctx: 2048 diff

Setting Ollama’s context window size

Ollama uses a 2k context window by default, which is very small for working with aider. Unlike most other LLM servers, Ollama does not throw an error if you submit a request that exceeds the context window. Instead, it just silently truncates the request by discarding the “oldest” messages in the chat to make it fit within the context window.

Except for the single 2k context result, all of the Ollama results above were collected with at least an 8k context window. An 8k window is large enough to attempt all the coding problems in the benchmark. Aider sets Ollama’s context window to 8k by default, starting in aider v0.65.0.

You can change the Ollama server’s context window with a .aider.model.settings.yml file like this:

- name: ollama/qwen2.5-coder:32b-instruct-fp16
  extra_params:
    num_ctx: 8192

Choosing providers with OpenRouter

OpenRouter allows you to ignore specific providers in your preferences. This can be used to limit your OpenRouter requests to be served by only your preferred providers.

Notes

This article went through many revisions as I received feedback from numerous members of the community. Here are some of the noteworthy learnings and changes:

  • The first version of this article included incorrect Ollama models.
  • Earlier Ollama results used the too small default 2k context window, artificially harming the benchmark results.
  • The benchmark results appear to have uncovered a problem in the way OpenRouter was communicating with Hyperbolic. They fixed the issue 11/24/24, shortly after it was pointed out.