PILLAR_CMP-LLMS LIVE CATEGORY LLMS · 4 ITEMS

Compare llms.

4 llms side-by-side on consistent axes. Per-axis values carry lastCheckedAt + confidence + source.

4 items tracked

▸ PICK ITEMS // LLMS

// AXIS	Claude Sonnet 4.5Default reasoning + tool-calling model	GPT-5Default reasoning + tool-calling on the OpenAI side
Pricing	$3 / $15 per M tokens (input / output) Input cache discount lowers effective cost on repeat calls. checked 2026-05-07 · medium confidence	$2.50 / $10 per M tokens (input / output) Reasoning effort settings can multiply token use 2-3x. checked 2026-05-07 · medium confidence
Context Window	200k checked 2026-05-07 · high confidence	256k checked 2026-05-07 · high confidence
Mcp Support	first-class Claude was the canonical MCP client; tool-calling is native.	via Apps SDK OpenAI's Apps SDK builds on MCP-style servers; native tool calling also works directly.
Leads At	Plain-English reasoning Following multi-step instructions Honest uncertainty	Massive consumer reach via ChatGPT Apps SDK distribution Image + audio capabilities
Lags At	Latency vs Haiku tier Image generation (none — text-only)	Reasoning settings can be opaque Fewer cross-vendor host integrations than Claude
Verdict	Default for non-trivial agentic work that lives outside ChatGPT.	Default if your audience lives in ChatGPT, or if you need image/audio out of the box.

Share this comparison: copy the URL — your selection is preserved.

OTHER CATEGORIES

DISCLOSURES

Microsoft pages disclose Sush's employer.
No vendor pays for placement.
Verdicts are scoped to a use-case (``default for X'' / ``use when Y''). Not winner / loser.
Every axis carries lastCheckedAt and confidence. Pricing in particular drifts — we date it explicitly.