Anthropic
Anthropic's largest frontier model — strongest at hard coding, agentic and reasoning work.
There is no open-weight model that matches Claude Opus outright. What the community reaches for instead are the strongest open coding and reasoning models — run locally, with no per-token cost and full data control. Closest by capability class:
These are community-regarded closest open-weight models by capability class — an editorial judgement, not a benchmark ranking. Sizes and hardware/speed figures are derived from the catalog and the same fit engine used across the site, judged at a 8,192-token context. Speeds are estimates, not measurements.
Community-regarded closest open coding models. The Qwen3-Coder family is the most widely cited local stand-in for Claude on code; the largest MoE variant is the closest in raw capability, the 30B-A3B variant the closest you can actually run on a single desktop GPU.
| Local model | Size | Hardware to run it | Est. speed |
|---|---|---|---|
| Qwen3-Coder-30B-A3B-Instruct Qwen tags: Coding, Long context | 30.5B (MoE) | Runs fully on 20 GB (RX 7900 XT) at IQ4_XS Runs fully on GPU @ IQ4_XS · 8K ctx | 184–246 tok/s (estimate) |
For Opus-style reasoning and long-form problem solving, the largest open generalists and reasoning-distill models are the community's closest local picks.
| Local model | Size | Hardware to run it | Est. speed |
|---|---|---|---|
| gpt-oss-120b openai | 120.4B (MoE) | Runs fully on 96 GB (RTX Pro 6000 Blackwell 96GB) at F16 Runs fully on GPU @ F16 · 8K ctx | 190–253 tok/s (estimate) |
| Qwen3.6-35B-A3B Qwen tags: Long context | 36B (MoE) | Runs fully on 24 GB (RTX 4090) at Q4_K_M Runs fully on GPU @ Q4_K_M · 8K ctx | 27–35 tok/s (estimate) |
| Qwen3-32B Qwen | 32.8B | Runs fully on 24 GB (RTX 4090) at Q4_K_M Runs fully on GPU @ Q4_K_M · 8K ctx | 28–37 tok/s (estimate) |
| DeepSeek-R1-Distill-Qwen-32B deepseek-ai | 32.8B | Runs fully on 24 GB (RTX 4090) at Q4_K_M Runs fully on GPU @ Q4_K_M · 8K ctx | 27–37 tok/s (estimate) |
"Hardware to run it" is the smallest single GPU in our roster that runs the model's best-fitting quant fully on GPU at this context, computed by the same fit engine used sitewide. "Est. speed" is a modelled generation-throughput range labelled estimate (D8), never a measurement. Equivalence is a conservative editorial call and no benchmark scores are implied.