Two models, side by side.
Pick two models. Same suite, same machine, the better value is marked.
So far one model has been tested. The comparison gets interesting as soon as the second one is in.
| Metric | Qwen3.8 Flash-NextQwen (Alibaba) |
|---|---|
| Speed | |
| Output at 1k (tokens/s) | 123.6 |
| Output at 32k (tokens/s) | 109.3 |
| First token at 1k (s) | 0.79 |
| First token at 32k (s) | 22.57 |
| Prefill at 32k (tokens/s) | 1,540 |
| 4 concurrent, total at 1k (tokens/s) | 217.5 |
| VRAM after loading (GB) | 93.6 |
| Tasks | |
| Coding passed | 4/5 |
| Agentic (average points) | 7.0/7 |
| Puzzles right | 5/5 |
| German: requirements met | 2/2 |
| Setup | |
| Quantization | UD-Q4_K_XL GGUF |
| Parameters | 125B parameters, 6B active |
| Tested on | 1 Oct 2026 |
Speed compared
Who runs the tests
Techat Lab is a project by Luca Hansmann. Luca also runs Techat, a platform with the AI assistant Techie, and this page promotes it. The measurements are independent of that: same suite, same machine.
The test machine
- GPU
- NVIDIA RTX PRO 6000 Blackwell, 96 GB VRAM
- CPU
- Intel Core i9-14900K
- RAM
- 64 GB
- Runtime
- llama.cpp (llama-server)
AI notice AI
The test suite collects the measurements automatically. Coding verdicts, conclusions and short texts are written with the help of Techie, Techat's AI assistant, from measurements and recordings. Such texts carry the “AI” mark.
Use these models in Techat
Techie, the AI assistant in Techat, normally runs on Qwen3.8 Flash-Next from this lab. It handles your website, email, bookings and contacts right in the chat.