Visit site
Auto-disqualified
Free tier is demo-only: no genuine product access before a sales call.
Overall score
About
SubQ is a sub-quadratic LLM built for long-context tasks, with a 12M-token context window aimed at agents that need to reason over full codebases or extended histories. Subquadratic ships the model behind an OpenAI-compatible API plus a coding CLI (SubQ Code) and a search tool (SubQ Search), all in private beta as of May 2026. The company raised $29M seed and claims 1/5 the cost of leading LLMs, though benchmarks remain unvalidated by independent researchers.
Best for: Engineering teams building agents that need to reason over entire codebases or long histories in a single prompt, and who are willing to wait for general availability before depending on the model in production.
Pricing
Private beta
- Monthly
- n/a
- Annual /mo
- n/a
- Billing
- flat
- Notes
- API access to the 12M-token SubQ model;SubQ Code CLI;SubQ Search;Streaming and tool use · Access via request form only; no public pricing disclosed. Vendor claims 1/5 the cost of leading LLMs but does not publish a per-token rate.
| Tier | Monthly | Annual /mo | Billing | Notes |
|---|---|---|---|---|
| Private beta | n/a | n/a | flat | API access to the 12M-token SubQ model;SubQ Code CLI;SubQ Search;Streaming and tool use · Access via request form only; no public pricing disclosed. Vendor claims 1/5 the cost of leading LLMs but does not publish a per-token rate. |
Key features
- 12M token context window
- OpenAI-compatible API
- Sub-quadratic sparse-attention architecture (~1,000x attention compute reduction at 12M tokens)
- 150 tokens per second inference
- SubQ Code CLI for coding agents
- SubQ Search for retrieval over long context
Trust & compliance
- Stage range
- Seed → Growth
- Founded
- 2025
- Status
- active
- SOC 2
- unknown
- GDPR
- unknown
- Data residency
- unknown
- External rating
- n/a
- Last verified
- May 2026
Reviews
Be the first to share your experience.