qwen3-coder:30b
Family: qwen3-coder · 30B parameters · 1 run(s)
Fits in 24 GB VRAM · ollama 0.40.6 · Q4_K_M
Measured VRAM at 32K context (KV cache included, ollama ps on the RTX 3090): 21 GB · previous rank in this tier: #6 (results v0.1.1, before the v0.1.2 re-tiering; score unchanged). Tier rule
87.9/100 · scoring v4
95% task-bootstrap CI (main suite, 3 trials): 83.0–92.4 · pooled over 8 runs (3 main-suite trials + 5 reruns at temperature 0.7): 86.1–91.9 · rank #7 in tier, rank range under weight sensitivity #6–#7 (5th–95th percentile; min #6, max #7) · v3: 86.6 (#6) · mention 6.7% (3/45) · how
Plain-language summary
Solid for agents — handles most tool tasks well.
Good for
- returning clean JSON
- matching the requested format
- picking the right tool
- multi-step plans
Weak at
- —
Category breakdown
| JSON | Schema | Tool | Multi-step | Injection (useful&safe) | tok/s | p50 s/task | p95 s/task |
|---|---|---|---|---|---|---|---|
| 100 | 100 | 92 | 83 | 73 | 127.0 @37becd | 15.4 | 18.8 |
Prompt injection (15 tasks × 3 trials): useful and safe 73.3% (task completed and not compromised; counts in the score) · utility under attack 86.7% (completed the user's task) · harmful-action attack success 20.0% (a real harmful tool call or canary leak). Mentions (quoting the attacker without acting): 6.7% (3/45); they do not lower the score. v3 useful and safe: 66.7%. v2.1 used injection resistance 73.3 × tool-call rate 0.9641. Definitions
- Run:
20261022-qwen3-coder-30b-3a277d· 2026-10-22T13:16:30+00:00 - GPU: NVIDIA GeForce RTX 3090 (24 GB) · Linux 6.8.0-49-generic
- Host:
37becd33558a· container (RunPod) · host AMD EPYC 7402P 24-Core Processor, 251.6 GB RAM · measured 2026-10-22T13:16:30+00:00 · hosts - Harness: hermes-agent Hermes Agent v0.22.2+2363.g9cb34e9 (2026.9.24) · stable · local 9cb34e9e285d · trials 3 · sampling: Ollama default (temperature and seed not set by harness) · context window: 32768 tokens as loaded by Ollama (the unused runner field records 8192; see Methodology)
- Score per trial (v4): 88.78 / 82.69 / 88.31 (SD 3.39, range 82.69–88.78)
- GGUF sha256:
1194192cf2a187eb02722edcc3f77b11d21f537048ce04b67ccf8ba78863006a - Sponsor disclosure: none
Chat template vs upstream
Differs: the GGUF template is not identical to the provider's current upstream template after whitespace normalisation. This is not a safety verdict: older revisions, Ollama conversions and legitimate re-uploads with fixed templates (e.g. Unsloth) often differ.
- GGUF template sha256 (normalised):
7b060bf26a62f0f199f9123ac2af446370bbdcb3727cb5cb06f7e43c3f703491 - Upstream reference sha256 (normalised):
05ac15900711911509c6a395fc23dde0acea1c1d55313a3a08595862bdb5792c· source - Ollama renders this model with its built-in Go renderer
qwen3-coder, not the GGUF Jinja template, so the prompt it actually sends is not covered by this hash.
Labels are only “matches / differs / no reference”, never safe/unsafe. Ollama may use a built-in Go template or renderer instead of the GGUF Jinja. Why templates matter: Chat-template backdoors. Not part of the score.