DeepSeek vs Sol
Open-weight cost disruptor vs closed coding ceiling — when is 96% SWE worth 30× the tokens? Public stats are peer-relative ranks (40–95 band among scored Builds), not absolute “% intelligence.”
DeepSeek V4 Pro · API · Max effort
The open ceiling — near-frontier coding without the closed-model tax.
GPT-5.6 Sol · API · Max effort
The ecosystem operator — max-effort Sol for hard agents, not cheap chat.
Six attributes (peer public)
Same-harness & loadout
| Signal | DeepSeek | Sol |
|---|---|---|
| SWE Verified | 77.4% Vals mini-swe · 80.6% vendor | 96.2% Vals mini-swe |
| GPQA Diamond | 88.8% AA (max) | 94.1% AA |
| HLE (closed-book) | 35.9% AA (max) | 47.2% AA |
| List price (in/out) | ~$0.44 / $0.87 | $5 / $30 |
| Open weights | Yes · MIT | No |
Large Existing Codebase
Decisive on Execution: Sol — public ranges do not overlap (DeepSeek 40.7–48.7 vs Sol 90.9–98.9).
- Large Existing CodebaseSol · conf A
Same-class agentic gap is large on independent boards; peer STR also Sol.
Quest Fit → - Autonomous Coding AgentSol · conf A
Ceiling when wrong is expensive — DeepSeek for volume/cost agents.
Quest Fit → - Best Under BudgetDeepSeek · conf A
Order-of-magnitude cheaper list price; watch max-effort verbosity.
Quest Fit → - Vibe Code a New AppDeepSeek (cost) / Sol (quality) · conf B
Pick Sol if quality is unconstrained; DeepSeek if spend is the constraint.
Quest Fit →
DeepSeek: Open frontier at ~$0.44/$0.87 — serious coding without Sol’s sticker or max-effort tax.
Sol: Independent SWE Verified peak of the peer set (96.2% mini-swe) — pay $5/$30 and ~151s TTFT at max.