Subquadratic launches SubQ Snap: a 384K-context decision model with constant compute per token at $0.05/M
ryanallred · x · 2026-10-13
Startup Subquadratic released SubQ Snap, a long-context "decision model" aimed at structured decisions inside enterprise agentic workflows, designed to offload simple agent steps from frontier LLMs.
Key claims:
- 384K multimodal context (12x Jev's 32K) with compute per token staying constant as context grows
- Scores 61.7 on Decision Index v0.2.1 vs Jev's 57.9, leading on 13 of 16 enterprise benchmarks
- Processes 100K tokens and returns a decision in under one second
- Priced at $0.05 per million input tokens with free output; first 10,000 sign-ups get 100M free tokens
The team also demoed Snap solving Sudoku flawlessly while GPT-Astra made 17 mistakes and Jev failed outright. Latency is still being tuned, and the company is soliciting compute offers to expand capacity.
More from Models
- SemiAnalysis: Anthropic A/B testing ~20% lower usage limits on some accounts — imjustnewatai · 2026-10-23
- Swift-1.5-Qwen3.8-27B ported to hipfire MQ4: 2.2x faster end-to-end than GGUF on a 7900 XT — Queasy_Asparagus69 · 2026-10-23
- Boundless offers abliterated DeepSeek V4.1 Flash for defensive security testing — MoonL88537 · 2026-10-23
- Anthropic uses Claude to formally verify security properties in 100+ open-source libraries at RISC-V assembly level — AnthropicAI · 2026-10-23
- Are models in an uncanny valley? Researcher warns users to stay vigilant — danielrock · 2026-10-23
- Uncensored Swift 1.5 Qwen 3.8 Flash Next Released as 4-bit GGUF with Strata-Ready Artifacts — jjusko20 · 2026-10-23