Model← Models
Qwen3.6 35B-A3B
35B mixture of experts, 3B active, MLX 4-bit (also held as UD-Q5_K_M GGUF), 19.0 GB.
- Parameters
- 35B / 3B active
- Quant
- MLX 4-bit (also held as UD-Q5_K_M GGUF)
- Size
- 19.0 GB
- Family
- qwen3.x
qwen3.6-35b
Measured, not seatedMoEresidentllama.cppmlxomlx
Model card ↗Outcome as of 20 September 2026, from the seat registry. The write-up below was made when the model was tested.
Why it was tested
Tested as the fast general MoE, and used as the daily coding driver in DSH. Moved to an MLX build on 2026-09-09.
What came of it
Held the speed record until 2026-09-15, when bonsai-8b-1bit measured 68.5 tok/s on a quiet box. Was: 66.4 tok/s single-stream and 193.2 aggregate at n=32, in 19GB, beating the GGUF build's 52.4. Its weakness is tool-seeking. It calls an announced tool 2/12 and an unmentioned one 0/12, so it needs if-then rules in tool descriptions to be usable agentically, and it goes to 12/12 when it gets them. Requires --tool-call-parser qwen3_coder_xml or its XML tool calls never become OpenAI tool_calls and it looks like it is ignoring every tool. In practice it drives opencode well: 75 tool calls, zero errors, because coding tools are ones the model genuinely cannot answer without.
| Suite | Score | 95% interval | Passed | Runtime · profile | Recorded |
|---|---|---|---|---|---|
| grounding | 82.4% | 74.8–88.1 | Passed 103/125 | llama.cppextraction | 2026-08-27 |
| nullschema | 61.1% | 49.6–71.5 | Passed 44/72 | llama.cppextraction | 2026-08-27 |
| screening | 86.7% | 75.8–93.1 | Passed 52/60 | llama.cppextraction | 2026-08-27 |
Runs marked unenforced were measured through mlx_lm on a suite that requires constrained decoding. mlx_lm accepts the schema and ignores it, so those numbers do not measure what the suite is designed to measure. They are kept and labelled rather than deleted.
Every run log entry for this model (5)
| Finished | Suite | Runtime · profile | Score | Coverage | Grader · dataset · config |
|---|---|---|---|---|---|
| toolcalltoolcall/1 | not recordedextraction | 65.0% | Coverage 60/60 | g 804e8fc9457649d8d 10af88c170513155c 8e5c69081ab588b0 | |
| nullschemanullschema/1 | llama.cppextraction | 61.1% | Coverage 72/72 | g d7448e0f5fc3f9afd 3b91280bb40372efc 7a652a6a84f42ef5 | |
| screeningscreening/1 | llama.cppextraction | 86.7% | Coverage 60/60 | g 0661b49a10be2d2ad 64c97ca2269c29afc 4246ff56e771b097 | |
| groundinggrounding/1 | llama.cppextraction | 82.4% | Coverage 125/125 | g 30b933dcc6a7e2d4d dcb396faf8b83b0ac 279f3a4bd1ff0c7e | |
| groundinggrounding/1 | llama.cppextraction | 86.3% | Coverage 73/73 | g 30b933dcc6a7e2d4d dcb396faf8b83b0ac 279f3a4bd1ff0c7e |
- Qwen3 Coder 30BMeasured, not seated
30B · MoE · Q4_K_M · 18.0 GB
The first coding model measured against a real coding contract rather than an extraction suite.
- Qwen3.5 122B-A10BRemoved
122B · MoE · IQ3 · 45.0 GB
Tested as the largest MoE that would fit, on the assumption that more parameters would win.
- Qwen3.8 27BSeated · agentic
27B · dense · MLX 4-bit (MTP build, template patched to reasoning_effort=medium) · 15.0 GB
The dense counterweight to the MoE fleet, and the model where MLX first beat llama.cpp. Promoted to the agent seat on 2026-09-09 after toolcall/1.
- Qwen3.8-Flash-NextRemoved
180B · MoE · AD-3.84bpw M64 · 79.1 GB
Tested because the M64 shard layout claims to run a 180B model on 64GB by keeping a 51B n-gram embedding table on SSD.