Expired50% confidenceBenchmarkExact time
On the main model 千问3.6 35B A3B, the 128G machine won seven of the nine levels, tied one, and lost one on the characters-generated-per-second metric, i.e., it was faster in 78% of the levels.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
11/1/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 11/15/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedIn a model comparison test, qwen2.5:3b correctly understood pace, produced 0 hallucinated place names, had a length error of 0.3%, peak memory of 259 MB, and took 106 seconds.72% similarUnverifiedThe Qwen 3.5 model with the compiled-in CPU got 39 of the same 40 multi-digit multiplication problems correct.70% similarUnverifiedThe performance gap between a carefully quantized 4-bit model and the original FP16 model is typically within 1–3 percentage points.68% similarUnverifiedQwen3 27B achieved a 100% pass rate in the 256K long-context memory retrieval test (needle in a haystack), with all 15 runs being accurate.68% similarUnverifiedQwen 3.6 27B outperformed Muse Glimmer in both the semantic three-run perfect accuracy rate and the format accuracy rate.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1122371API
curl https://kongchang.com/api/v1/knowledge/claims/1122371MCP
get_claim(id=1122371)