Model benchmarks
qwen3.8-flash-next-coder-iq1_m local LLM performance
As of November 2026, qwen3.8-flash-next-coder-iq1_m runs at up to 172.3 tok/s for local inference (best of 13 community benchmark runs across 7 GPUs).
Model size
Unknown
Peak speed
172.3 tok/s
Average speed
77.1 tok/s
Avg PP
531.4 tok/s
Min memory
3.4 GB
Max context
262,144 tokens
Avg output / run
76,749 tokens
Avg runtime / run
21m 44s
Avg quality
79.9
Benchmark runs
13
GPUs tested
7
Quality by task
Average LLM-judged quality (0–100) with the run-to-run spread shown as a P5–P95 band, overall and for each benchmark task, across all 13 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 66.7 | 79.9 | 87.1 |
| Agent Workflow | 80.8 | 87.3 | 92.5 |
| Code Generation | 24.2 | 62.1 | 83.4 |
| Role Play & Narrative | 48.1 | 83.5 | 93.6 |
| Research & Analysis | 80.5 | 86.7 | 90.4 |
Performance by hardware and tool
Every hardware/tool/quantization combination qwen3.8-flash-next-coder-iq1_m has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 5090 | Strata | — | 172.3 tok/s | 172.3 tok/s | 87.0 GB | 131,072 tokens | 80.8 | 1 |
| NVIDIA GeForce RTX 3090 | Strata | — | 104.3 tok/s | 90.3 tok/s | 48.5 GB | 65,536 tokens | 81.7 | 4 |
| NVIDIA GeForce RTX 4090 | OpenAI-compatible | — | 85.1 tok/s | 85.1 tok/s | 88.2 GB | 256,000 tokens | 86.9 | 1 |
| AMD Radeon RX 7900 XTX | Strata | — | 81.4 tok/s | 70.1 tok/s | 3.4 GB | 131,072 tokens | 77.8 | 3 |
| AMD Radeon RX 9070/9070 XT/9070 GRE | Strata v0.2.6 | — | 62.0 tok/s | 62.0 tok/s | 43.7 GB | 65,536 tokens | 86.4 | 1 |
| AMD Radeon RX 9070/9070 XT/9070 GRE | Strata | — | 43.3 tok/s | 43.3 tok/s | 3.4 GB | 262,144 tokens | 84.0 | 1 |
| AMD Radeon RX 7800 XT | Strata | — | 42.2 tok/s | 42.2 tok/s | 34.5 GB | 65,536 tokens | 63.8 | 1 |
| AMD Radeon RX 9070 XT | llama.cpp | — | 25.6 tok/s | 25.6 tok/s | 31.3 GB | 65,536 tokens | 76.6 | 1 |
Benchmark runs
All 13 qwen3.8-flash-next-coder-iq1_m runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is qwen3.8-flash-next-coder-iq1_m good for coding?
- In our benchmarks, qwen3.8-flash-next-coder-iq1_m scores 62.1/100 for coding. It runs at about 77.1 tok/s, so if you want more speed, ornith-1.5-35b-a3b-apex-mtp-i-quality is faster (~105.2 tok/s) and still scores well for coding (83.6/100).
- Is qwen3.8-flash-next-coder-iq1_m good for agentic (tool-using) tasks?
- In our benchmarks, qwen3.8-flash-next-coder-iq1_m scores 87.3/100 for agentic workflows. It runs at about 77.1 tok/s, so if you want more speed, ornith-1.5-35b-a3b-apex-mtp-quality is faster (~103.2 tok/s) and still scores well for agentic workflows (88.5/100).
- How fast is qwen3.8-flash-next-coder-iq1_m for local inference?
- Across 13 community benchmark runs, qwen3.8-flash-next-coder-iq1_m reaches up to 172.3 tok/s and averages 77.1 tok/s, with the fastest results on NVIDIA GeForce RTX 5090.
- How much memory does qwen3.8-flash-next-coder-iq1_m need?
- The leanest observed configuration used about 3.4 GB of memory.
- Which tools have been used to run qwen3.8-flash-next-coder-iq1_m?
- Benchmarks were submitted using OpenAI-compatible, Strata, Strata v0.2.6, llama.cpp. Results are community-contributed and updated as new runs arrive.