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

Model datasheet · gpt-oss

gpt-oss-120b

The big sibling of gpt-oss-20b, and the same bargain at four times the size: shipped natively in MXFP4, so there is no quant ladder to choose from — one 63GB file is the model. Reasoning and tool use are excellent; world knowledge is not. It was built to look things up.

Vendor
OpenAI
Architecture
MoE · 117B total / ~5.1B active
Context
131,072 tokens
License
Apache-2.0
Released
2025-08-05
Vision
No

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

§1 Characteristics by quantization band

Q4–Q5 — 4–5 bit

63.4GB weights · e.g. mxfp4 + KV: 293MB at 8K · 1.1GB at 32K · 4.5GB at 128K
Math & reasoning 9/10
data checked Aug 2026
Coding 7/10
data checked Aug 2026
  • vendor-model-cardSWE-bench Verified = 62.4 @ mxfp4 · vendor-reported · 2026-08-31

    Codeforces Elo 2463, Aider Polyglot 44.4. Only 1.7 points above the 20b despite 5.6x the parameters; the Aider figure is the more sober read of everyday editing.

Tool calling / agentic 8/10
data checked Aug 2026
  • vendor-model-cardTau-Bench Retail = 67.8 @ mxfp4 · vendor-reported · 2026-08-31

    Against the 20b's 54.8. Tool use is the family's organizing principle — browsing, python and functions are first-class in the harmony format.

Factuality 3/10
data checked Aug 2026
  • vendor-model-cardSimpleQA = 16.8 @ mxfp4 · vendor-reported · 2026-08-31

    PersonQA hallucination rate 49.1% — OpenAI's own number. Better than the 20b's 6.7 / 53.2% and still poor in absolute terms. HealthBench 57.6.

§2 Known issues & what fixes them

ISSUE-01 severe Q4–Q5 · factuality

Hallucinates confidently on world knowledge. SimpleQA 16.8 and a 49.1% PersonQA hallucination rate are OpenAI's own published figures — better than the 20b's, and still bad enough that this model should not be trusted as a source of facts. It is built to reason and call tools, not to remember.

Fix
RAG (retrieval) [strong]

Give it documents to read. Retrieval turns the weakness into a non-issue for document work.

Fix
MCP / external tools [strong]

Wire up search. The model is explicitly designed to look things up rather than recall them.

Evidence · 2 sources · community-consensus
ISSUE-02 mild Q4–Q5 · coding, math

There is no quant choice to make: the MoE weights were post-trained in MXFP4, so the 63GB native file is the model rather than a compression of it. Convenient — no band to agonise over — but the usual escape hatch is missing. If 63GB does not fit, there is no smaller official build, and requantizing an already-4-bit model is not the same operation as quantizing from bf16.

No fix
No known fix — choose a different model

Use gpt-oss-20b instead if 63GB is out of reach; same architecture, 12GB.

Evidence · 2 sources · community-consensus