Quick Answer
Is Gemini 4 Available Yet? Carbon and Argon vs Claude and GPT
The short answer
Yes — both Gemini 4 models are out. As of November 12, 2026:
- Gemini 4 Argon — generally available since October 13, 2026 on the paid Gemini API, Gemini Enterprise, Antigravity and the Gemini app (AI Pro/Ultra).
- Gemini 4 Carbon — Google’s strongest model, first given to cyber defenders on October 28, then opened on November 10, 2026 to all paying Gemini API developers (AI Studio, Android Studio), Gemini for business, and Google AI Ultra subscribers in the Gemini app, AI Mode and Antigravity.
Neither runs on the free API tier. Both are $2/$10 per million tokens until November 30, 2026, then $4/$20.
Gemini 4 vs Claude vs GPT: price and positioning
Standard API prices per million tokens, from each vendor’s pricing page (November 12, 2026).
| Model | Input / output now | From Dec 1, 2026 | Context | Best at (vendor claims) |
|---|---|---|---|---|
| Gemini 4 Carbon | $2 / $10 | $4 / $20 | 1,048,576 (unified in+out) | Long multi-step coding, cybersecurity, research |
| Gemini 4 Argon | $2 / $10 | $4 / $20 | 1,048,576 | Most coding and knowledge work, fewer tokens |
| Claude Opus 5.5 | $4 / $20 | unchanged | 1M, 128K out | Anthropic’s recommended daily driver; leads OSWorld-2.1 in Google’s table |
| Claude Fable 5.2 | $8 / $40 | unchanged | 1M, 128K out | Anthropic’s top model (released Nov 9) |
| GPT-6 Astra | $10 / $50 | unchanged | 1,050,000 | OpenAI flagship; led FrontierSWE 2 in Google’s Oct 28 table |
| GPT-6.1 Sol | $2 / $10 | unchanged | 1,050,000 | OpenAI’s price-performance pick |
After December 1, Carbon and Argon cost exactly what Claude Opus 5.5 costs per token. The real difference is tokens per task: Google says Carbon “uses more tokens than Argon to explore thoroughly”, so it will cost more per job than Argon at the same rate. All prices: current API prices.
How good is Carbon?
Google’s claims, which independent testing has not yet fully confirmed:
- #1 on Artificial Analysis Intelligence Index v5.0 with 59 against 58 for Claude Opus 5.5 (max) and 50 for Argon (October 28 announcement). v5.0 scores are not comparable with older index versions.
- #1 on CursorBench 4.0, Perplexity’s WANDR and Zapier’s AutomationBench, at “more than 2X lower cost per task compared to the next best external model” (November 10).
- #2 on Artificial Analysis’s new Multimodal Knowledge Agent benchmark, at more than 6x lower average cost per task than the leader.
- Not first everywhere: Google’s own launch table had GPT-6 Astra ahead on FrontierSWE 2 and Claude Opus 5.5 ahead on OSWorld-2.1.
Background: what is Gemini 4 Carbon and Gemini 4 Argon vs Claude Opus 5.5 vs GPT-6 Astra.
Which should you use?
- Hardest agentic coding, security review, research: Gemini 4 Carbon, Claude Fable 5.2 or GPT-6 Astra. Carbon is the cheapest per token of the three, even after December 1.
- Everyday coding and knowledge work: Gemini 4 Argon, Claude Opus 5.5 or GPT-6.1 Sol. On price, GPT-6.1 Sol stays at $2/$10 after Gemini doubles.
- Committed to Google Cloud or Android: Argon by default, Carbon for the jobs Argon fails.
Limits worth knowing
- Implicit caching only — Google does not support explicit context caching on Carbon or Argon; cache reads are $0.10 now, $0.20 from December 1.
- Thinking is billed as output, and Carbon’s thinking levels are low, medium and high.
- Max output is 65,536 tokens by default, up to 1,048,576 with long decode.
- No Live API, image or audio generation on Carbon.
Last verified: November 12, 2026.