Google's newest flagship launched with an introductory rate that runs through November 30, 2026. After that, its list price lands exactly on Claude Opus 5.5's — so any cost comparison depends on when you read it.
Today Gemini 4 Argon is half the price of Claude Opus 5.5 ($2.00/$10.00 vs $4.00/$20.00 per 1M). From December 1 Argon doubles to the same input, output and cache-hit rates as Opus 5.5, so a long-term choice should rest on quality, tooling and caching mechanics rather than the intro discount.
| Claude Opus 5.5 | Gemini 4 Argon | |
|---|---|---|
| Provider | Anthropic | |
| Tier | Flagship | Flagship |
| Input / 1M tokens | $4.00 | $2.00 |
| Output / 1M tokens | $20.00 | $10.00 |
| Cache hit / 1M | $0.20 | $0.10 |
| Cache write / 1M | $5.00 | — |
| Context window | 1M | 1M |
| Notes | New leading Opus model (Sep 22, 2026); Fable 5.1-level performance at 20% less than Opus 5; included on Pro/Max/Team at no extra cost | New top frontier model, generally available Oct 13, 2026 for AI Pro/Ultra subscribers and via the API (model id gemini-4-argon); long-decode generation up to 1M tokens; intro price thru Nov 30, 2026, then $4/$20 (cache read $0.20); not available on the free tier |
Gemini 4 Argon is 2.0× cheaper on input tokens ($2.00 vs $4.00 per 1M). Gemini 4 Argon is 2.0× cheaper on output tokens ($10.00 vs $20.00 per 1M).
| Workload | Claude Opus 5.5 | Gemini 4 Argon | Difference |
|---|---|---|---|
| Support chatbot 10,000 conversations × (2,000 input + 400 output tokens) | $160 | $80.00 | Gemini 4 Argon saves $80.00 |
| Document Q&A 1,000 questions × (30,000 input + 800 output tokens) | $136 | $68.00 | Gemini 4 Argon saves $68.00 |
| Coding agent session 4M input tokens (90% cache hits) + 150,000 output tokens | $5.32 | $2.66 | Gemini 4 Argon saves $2.66 |
Standard rates per 1M tokens from data on this site. The coding-agent row assumes 90% of input is read from the prompt cache (full input price where a model has no cache rate). Cache-write fees, batch discounts and long-context surcharges listed in the notes are not included. Updated November 2, 2026.
you want explicit control over prompt caching (named cache breakpoints with 5-minute or 1-hour lifetimes) and a price that is not about to change, or your evals favour Claude on your tasks.
you can benefit from the introductory rate during November 2026, you are already on Google Cloud, or your workload leans on Argon's long-decode generation of up to 1M tokens.
Google's pricing page lists Argon's intro rate through November 30, 2026, and double that from December 1 — which puts input, output and cache hits at exactly Opus 5.5's $4.00, $20.00 and $0.20. If you are sizing a budget for 2027, model it at parity. The workload table below reflects today's intro rate, so it overstates Argon's long-run advantage.
Argon uses implicit caching only: Google applies the cache-hit rate automatically when it detects a repeated prefix, and lists no cache-write fee. Opus 5.5 uses explicit caching — you mark what to cache and pay $5.00 per 1M tokens to write it (5-minute TTL; a 1-hour TTL costs 2× input). Implicit caching is simpler to adopt; explicit caching makes hit rates predictable, which matters when you are budgeting an agent that re-reads the same codebase hundreds of times.
Neither model adds a long-context surcharge: Anthropic bills the full 1M window at standard rates, and Google lists no higher tier for Argon. Both offer a 50% Batch discount, so for overnight jobs the ratio stays the same. Note that Argon is not available on the Gemini API free tier, so prototyping costs from the first call.
Only until November 30, 2026. From December 1 Google's list price doubles, which matches Opus 5.5 on input, output and cache-hit rates.
Google lists only implicit caching for Argon, with a cache-hit rate of $0.10 during the intro period and no separate write fee. Claude Opus 5.5 charges $5.00 per 1M tokens to write a 5-minute cache.
Not through the API: Google's pricing page says Argon is not available on the free tier.
Both list 1M tokens, and neither charges a long-context surcharge.
How we choose: prices come from each vendor's official pricing page and are re-checked every week. We take no affiliate fees, so rankings aren't for sale.