Known-issues sheets for local language models Rev 0.1.0 · reviewed every 7 days

Model datasheet · Qwen3.5

Qwen3.5 397B-A17B

The largest model on this site and the top of the Qwen3.5 line. Gated DeltaNet plus sparse MoE: only 15 of its 60 layers are full attention, so 256K of context costs about 8GB of cache — remarkably cheap for its size. The weights are the problem, not the cache.

Vendor
Alibaba
Architecture
MoE · 397B total / ~17B active
Context
262,144 tokens
License
Apache-2.0
Released
2026-02-16
Vision
Yes

→ GGUF (llama.cpp / LM Studio / Ollama)

§1 Characteristics by quantization band

Q2–Q3 — 2–3 bit

177GB weights · e.g. IQ2_M , Q3_K_M + KV: 240MB at 8K · 960MB at 32K · 3.8GB at 128K
Coding 6/10
data checked Aug 2026
  • quant-degradation-community — editorial band derivation @ Q3_K_M · aggregated · 2026-08-31

    Editorial derivation: two points down. At 3-bit a MoE this large loses expert-routing precision first, which surfaces as dropped tool arguments and broken multi-file edits before prose visibly suffers.

Tool calling / agentic 5/10
data checked Aug 2026
  • quant-degradation-community — editorial band derivation @ Q3_K_M · aggregated · 2026-08-31

    Editorial derivation: two points down. At 3-bit a MoE this large loses expert-routing precision first, which surfaces as dropped tool arguments and broken multi-file edits before prose visibly suffers.

Q4–Q5 — 4–5 bit

244GB weights · e.g. Q4_K_M + KV: 240MB at 8K · 960MB at 32K · 3.8GB at 128K
Coding 8/10
data checked Aug 2026
  • quant-degradation-community — editorial band derivation @ Q4_K_M · aggregated · 2026-08-31

    Editorial derivation: one point down from the measured bf16 figure. 4-bit is the band a model this size is actually run at, and the loss is real but modest.

Math & reasoning 8/10
data checked Aug 2026
  • quant-degradation-community — editorial band derivation @ Q4_K_M · aggregated · 2026-08-31

    Editorial derivation: one point down from the measured bf16 figure. 4-bit is the band a model this size is actually run at, and the loss is real but modest.

Tool calling / agentic 7/10
data checked Aug 2026
  • quant-degradation-community — editorial band derivation @ Q4_K_M · aggregated · 2026-08-31

    Editorial derivation: one point down from the measured bf16 figure. 4-bit is the band a model this size is actually run at, and the loss is real but modest.

Q6–Q8 — 6–8 bit

327GB weights · e.g. Q6_K + KV: 240MB at 8K · 960MB at 32K · 3.8GB at 128K
Coding 9/10
data checked Aug 2026
  • quant-degradation-community — editorial band derivation @ Q6_K · aggregated · 2026-08-31

    Editorial derivation: no drop from the measured bf16 figure. 6-8 bit is effectively lossless on a model this size — the differences that show up at 4-bit and below are not measurable here. Recorded explicitly rather than left to band fallback, which would otherwise borrow the 4-bit number and subtract a point.

Factuality 8/10
data checked Aug 2026
  • quant-degradation-community — editorial band derivation @ Q6_K · aggregated · 2026-08-31

    Editorial derivation: no drop from the measured bf16 figure. 6-8 bit is effectively lossless on a model this size — the differences that show up at 4-bit and below are not measurable here. Recorded explicitly rather than left to band fallback, which would otherwise borrow the 4-bit number and subtract a point.

Instruction following 9/10
data checked Aug 2026
  • quant-degradation-community — editorial band derivation @ Q6_K · aggregated · 2026-08-31

    Editorial derivation: no drop from the measured bf16 figure. 6-8 bit is effectively lossless on a model this size — the differences that show up at 4-bit and below are not measurable here. Recorded explicitly rather than left to band fallback, which would otherwise borrow the 4-bit number and subtract a point.

Math & reasoning 9/10
data checked Aug 2026
  • quant-degradation-community — editorial band derivation @ Q6_K · aggregated · 2026-08-31

    Editorial derivation: no drop from the measured bf16 figure. 6-8 bit is effectively lossless on a model this size — the differences that show up at 4-bit and below are not measurable here. Recorded explicitly rather than left to band fallback, which would otherwise borrow the 4-bit number and subtract a point.

Tool calling / agentic 8/10
data checked Aug 2026
  • quant-degradation-community — editorial band derivation @ Q6_K · aggregated · 2026-08-31

    Editorial derivation: no drop from the measured bf16 figure. 6-8 bit is effectively lossless on a model this size — the differences that show up at 4-bit and below are not measurable here. Recorded explicitly rather than left to band fallback, which would otherwise borrow the 4-bit number and subtract a point.

FP16 — full precision (fp16/bf16)

793GB weights · e.g. bf16 + KV: 240MB at 8K · 960MB at 32K · 3.8GB at 128K
Coding 9/10
data checked Aug 2026
Factuality 8/10
data checked Aug 2026
  • vendor-model-cardMMLU-Pro = 87.8 @ bf16 · vendor-reported · 2026-08-31

    No hallucination benchmark published; this is a knowledge score, not a truthfulness one.

Instruction following 9/10
data checked Aug 2026
  • vendor-model-cardIFEval = 92.6 @ bf16 · vendor-reported · 2026-08-31

    IFBench 76.5. Both are program-checkable, so there is little interpretive room.

Math & reasoning 9/10
data checked Aug 2026
  • vendor-model-cardAIME26 = 91.3 @ bf16 · vendor-reported · 2026-08-31

    GPQA 88.4, MMLU-Pro 87.8 — the highest GPQA figure in this catalogue.

Tool calling / agentic 8/10
data checked Aug 2026
  • vendor-model-cardBFCL-V4 = 72.9 @ bf16 · vendor-reported · 2026-08-31

    Terminal Bench 2 52.5 — the best terminal-agent figure here, though BFCL trails GLM-4.7's Tau2 result.

§2 Known issues & what fixes them

ISSUE-01 moderate Q2–Q3 / Q4–Q5 / Q6–Q8 / FP16 · long-context

The unusual shape of this model is that context is cheap and weights are not. Three of every four layers use linear attention, so the KV cache is only 30KB per token — a full 256K context costs about 8GB, less than GLM-4.7 spends on 32K. But the 4-bit weights are 244GB. Nothing about a long context will stop you here; holding the model will.

No fix
No known fix — choose a different model

Architectural, and mostly good news. Size the machine for weights and treat context as close to free.

Evidence · 2 sources · community-consensus
ISSUE-02 moderate Q2–Q3 / Q4–Q5 / Q6–Q8 / FP16 · coding, math

Every number scored here is Alibaba's own. Qwen3.5-397B is a flagship release with headline figures that beat much larger closed models, and no independent replication of those figures exists in this dataset. Treat them as claims. Its factuality and long-context cells are blank rather than estimated because nobody has published a hallucination or RULER result for it.

No fix
No known fix — choose a different model

Nothing to work around — this is a caveat on the evidence, not on the model.

Evidence · 2 sources · community-consensus