lifeidle / qwen3.8-flash-next-5090-laptop-24gb-strata
git clone https://github.com/lifeidle/qwen3.8-flash-next-5090-laptop-24gb-strata.git
Qwen3.8-Flash-Next 177B MoE (IQ3_S 3.44bpw) on RTX 5090 Laptop 24GB: full-model quality @ 97.7 tok/s · dual-lane 256K full-context unlocked on Strata 0.1.50 · 486K tokens in flight · 60+ measured rounds, 5 quant tiers · 中英双语实测实录
+0
stars, 24h · #71,058
+3
stars, 7 days · #33,611
+~11
stars, 30 days · #38,545
28
★ total · +3% this week
2
forks · 0 open issues
Star history (as recorded here)
LanguagePython
LicenseMIT
Created2026-09-10 (1mo ago)
Last push11d ago
OwnerUser
GitHub topics
256k-contextflash-nextggufiq3-xxslocal-llmmoemtpquantizationqwen3qwen38reasoningrtx5090self-hostedspeculative-decodingstratavision-language-model
Links
- GitHub: github.com/lifeidle/qwen3.8-flash-next-5090-laptop-24gb-strata
- Homepage: https://chat.specul.com
- Releases · Issues · Hacker News mentions
README
Loading README…