Mistral Large 4 vs Qwen 3.8 Flash
Compare models
Qwen 3.8 Flash is more worth it — 74% vs 50%
How we score worth-it? →European open-weights flagship with aggressive preview pricing.
- High one is actually much better — the feet connect to the pedals, no hub cap, relatively proper position.— hn
- Moving a lot of our stuff over to Mistral. Very much happier keeping data in Europe rather than leeching it all over the US.— hn ↗
- Mistral has been great for us as a provider for our EU-based company... since they added more features and improved their own model, we are quite happy.— hn ↗
- The dense graph pins at 16.1 GiB, and the model runs on a single 32 GB consumer card. Thank you for shipping the reference modeling code — five details are easy to get wrong and all of them change the answers.— huggingface ↗
- Mistral is good. "Only" 250 Euro, but GLM is cheap af, and mistral has absolutely zero of these weird-ass "5h limits" that became the defacto standard for some reason.— hn ↗
- Still too expensive and very slow when thinking. We need one more generation from Mistral and 50% of this already 'slashed' price to be competitive.— hn ↗
- I tried mistral 4 when it came out and it honestly felt like going back in time 6 months — my agents were making all kinds of mistakes I'd forgotten agents used to make.— hn ↗
- The benchmark numbers are dreadful and the cost is higher than Sol. There are a ton of models that are a lot smarter for a lot less money.— hn ↗
- Mistral 4 IMHO is not that good. I ran the full NVFP4 weight and couldn't stomach it after 3 days.— hn ↗
- When I last tried Mistral Large 4 for a code documentation task, it ignored my SKILL.md files and it had a few hallucinations. I wondered if I had failed to prompt it properly or if there was some problem with my harness.— hn ↗
“High-effort mode renders the pelican correctly — feet connect to pedals”
“Far from AGI — it just regenerated a well-known benchmark image”
“Aggressive preview pricing at $0.68/$2.09”
Alibaba's fast, cheap open-weight multimodal model (Qwen 3.8 family) with 1M context, deep thinking, and agentic coding — the budget Flash sibling to Qwen 3.8 Max.
- Qwen 3.8 Flash Next has only 6B activated parameters — on a single PRO 6000 it pre-fills over 10K tokens and decodes at 150 tps, with room to optimize. It really is sparse and fast.— v2ex ↗
- Qwen 3.8 flash next is way better than 27B. It's so good I don't even use Claude anymore.— hn
- Qwen Flash Next not only seemed less delusional than the 27B, but also made useful observations and problem-solved ways around many website access issues.— hn ↗
- Qwen3.8-next-flash is impressive though, especially with Strata or Basalt as inference engines. On a 5090 laptop with 64GB RAM one can run NVFP4 quants at 120-140t/s decode and 3000-4000t/s prefill.— hn ↗
- Qwen 3.8's Flash Next at A6B seems about juuuuust right— hn ↗
“Sparse and fast — only 6B active parameters, decoding at 150 tps on a single consumer GPU.”
“Flattened the VRAM barrier — a data-center-grade long-context model that runs on a consumer RTX 4090.”
“Superior quality for coding, and notably good at finding issues in generated code.”
Which should you pick?
The gap is clear: Qwen 3.8 Flash sits at 74% while Mistral Large 4 is at 50%, a 24-point difference. If you just want the safer pick, start with Qwen 3.8 Flash.
On entry price, Qwen 3.8 Flash is cheaper ($0.11 vs $0.7, Mistral Large 4 costs about 536% more).
Both are in the Chat category with similar coverage; the real difference is stability and cost — run one of your own tasks through each before deciding.
What users actually said about Qwen 3.8 Flash: "Qwen 3.8 Flash Next has only 6B activated parameters — on a single PRO 6000 it pre-fills o…"
FAQ
Is Mistral Large 4 or Qwen 3.8 Flash more worth it?
Qwen 3.8 Flash is more worth it, with a 74% worth-it score vs 50% for Mistral Large 4.
How much does Mistral Large 4 cost vs Qwen 3.8 Flash?
Mistral Large 4: $0.70/$2.10 per 1M · Qwen 3.8 Flash: $0.11 / $0.38 per 1M tokens.
What do users say about Mistral Large 4 vs Qwen 3.8 Flash?
Mistral Large 4: High-effort mode renders the pelican correctly — feet connect to pedals. Qwen 3.8 Flash: Sparse and fast — only 6B active parameters, decoding at 150 tps on a single consumer GPU..