MODEL VERSIONS ON THIS PAGE HAVE MOVED — CHECKED 8 OCTOBER 2026
Current line, with published API pricing per million tokens:
●
OpenAI — GPT-6 Sol (paid ChatGPT), GPT-6 Luna ($0.10 / $0.50), GPT-6.1 Sol ($2.00 / $10.00), GPT-6 Astra
●
Anthropic — Claude Opus 5.5, Sonnet 5.5 ($2.00 / $10.00), Haiku 5.5 ($0.10 / $0.50 below 100K), Fable 5.1
●
xAI — Grok 4.3 ($1.25 / $2.50), Grok 4.7 ($2.00 / $6.00)
●
Google — Gemini 3.5 Flash-Lite ($0.30 / $2.50), Gemini 3.8 Flash ($0.75 / $3.75 to 31 Dec), Gemini 4 Argon
The comparison below is kept because the reasoning still holds — how these tiers trade price against capability has not changed. The version numbers and prices have.
Current pricing table →
ACTION REQUIRED — AUGUST 31, 2026 DEADLINE
● Sonnet 5 now (through Aug 31): $2/$10/M — 60% cheaper than Opus 5
● Sonnet 5 after Aug 31: $3/$15/M — 40% cheaper than Opus 5
● Opus 5 (unchanged): $5/$25/M
● What to do before Aug 31: Run high-volume workloads on Sonnet 5, batch any large processing jobs, build and cache any high-cost context patterns
● What changes Sept 1: Sonnet 5 becomes $3/M — still cheaper than Opus 5 but the gap narrows from 60% to 40%
● When to use Opus 5 over Sonnet 5: Hard reasoning (ARC-AGI-3 30.2%), effort toggle (5-level), 128K output, Claude Max default
The intro pricing window closes August 31. If you have batch processing jobs, large-scale content generation, or any workload that can be front-loaded before the deadline, running them through Sonnet 5 at $2/M before September 1 saves 33% compared to the $3/M price after (or 60% compared to Opus 5). After the deadline: Sonnet 5 remains the right choice for writing, coding, and analysis that does not specifically need Opus 5's effort toggle or ARC-AGI-3 reasoning performance. The 40% price gap is still material for high-volume workloads.
Before August 31 — route to Sonnet 5: Content generation at volume, code review queues, batch document processing, RAG pipeline calls, any workload where quality difference from Opus 5 is marginal.
Use Opus 5 regardless of deadline: Tasks needing the effort toggle, extended reasoning (ARC-AGI-3 30.2%), very long outputs (128K), auto-fallback replacing hard refusals, and Claude Max sessions where Opus 5 is the default.
Last updated July 30, 2026. Related: Claude Opus 5 full review →