Flash model size chart puts Qwen3.8-Flash-Next smallest at 176B total parameters
BerryBlush_Rogers691 · reddit · 2026-11-18
A Reddit user discusses a local-AI size comparison chart of "Flash" class models:
- Qwen3.8-Flash-Next is the smallest at 176B (125B main model + 51B n-gram embeddings), followed by Hy4 Flash (224B), GLM-5.3-Flash (320B), MiniMax-M3.1-Flash (428B), Ling-3.1-Flash (560B), and DeepSeek-V4.1-Flash (748B).
- The bars aren't VRAM requirements: they include n-gram/Engram tables, and Qwen's table can be offloaded to system RAM — how parameters are stored matters a lot for local deployments.
- For Qwen4-Flash, the author wants better tool use and fewer dumb mistakes at similar hardware requirements; the bad outcome would be a much better model the community ignores because 3.8 already runs comfortably.
More from Infra
- TokScale leaderboard: #1 user burned 4.65T tokens, est. $2.43M in cost — olcan · 2026-11-18
- Inside Microsoft's massive Outlook outage: what broke and why fixes took so long — tomwarren · 2026-11-18
- vLLM ships DSpark draft model for Kimi K3.1 Pro with up to 30% higher acceptance length — vllm_project · 2026-11-18
- Wonderful runs in-house voice models on Together AI at 274ms first sentence — togethercompute · 2026-11-18
- Anthropic's S-1 Shows Only 25% of Compute Goes to Inference — Sparking a Bear Case — toptickcrypto · 2026-11-18
- LanceDB Raises $57.5M Series B for AI-Native Multimodal Data Stack — pzakin · 2026-11-18