// model hub overview
Gemini 4 Argon: API Pricing, Batch and Caching
Gemini 4 Argon is Google's foundation model featuring a context window of 1,048,576 tokens. Google positions it for long, multi-step work in software engineering, enterprise tasks in legal and finance, and cyber defense. The model is available via paid API tiers and select subscription plans.
API Pricing and Token Rates
| model | input / 1M | output / 1M | context |
|---|---|---|---|
| Gemini 4 Argon | $2 | $10 | 1,048,576 |
Rates come from the OpenRouter price list and are checked daily (prices updated 2026-11-11)
Competitor Cost and Context Comparison
| model | input / 1M | output / 1M | context |
|---|---|---|---|
| Gemini 4 Argon | $2 | $10 | 1,048,576 |
| GPT-6 Astra | $10 | $50 | 1,050,000 |
| Claude Opus 5.5 | $4 | $20 | 1,000,000 |
| Claude Fable 5.1 | $10 | $50 | 1,000,000 |
| GPT-6.1 Sol | $2 | $10 | 1,050,000 |
| Muse Wonder | $4 | $20 | 1,048,576 |
Across compared alternatives, Gemini 4 Argon and GPT-6.1 Sol share the cheapest output rate at $10 per million tokens, while GPT-6 Astra and Claude Fable 5.1 are the most expensive at $50 per million output tokens.
Who can use Gemini 4 Argon since October 13, 2026
- Developers and companies: the paid tier of the Gemini API (Google AI Studio, model id gemini-4-argon) and the Gemini platform for business. There is no free API tier for this model.
- Google AI Ultra and paid Google AI Pro plans: Google Antigravity (agentic coding). Not included in the free plan or Google AI Plus.
- Google AI Pro and Ultra subscribers: the Gemini app and AI Mode in Search. Google started the rollout on October 13 - the model may not show up in every Pro account on day one.
- OpenRouter: google/gemini-4-argon at the same introductory price.
- Cyber defenders can apply to the Fairwind Program; Google also offers 10,000 scientists the top Antigravity tier free for 6 months (eligibility and capacity apply).
Gemini API pricing per 1M tokens (USD)
| until Nov 30, 2026 | from Dec 1, 2026 | |
|---|---|---|
| Input | $2 | $4 |
| Output | $10 | $20 |
| Cached input | $0.10 | $0.20 |
| Batch input | $1 | $2 |
| Batch output | $5 | $10 |
Official Gemini API price list. Output includes thinking tokens.
Benchmarks according to Google
Figures published by Google - the vendor's own measurements, not independent tests.
| DeepSWE v1.1 | 77.9% |
| AutomationBench | 51.3% (#1) |
| LVBench | 91.7% |
| CWE-bench v1 | 68% (#1 =) |
| ERQA | 87% |
| Vals Index | #1 |
| Vals Finance Agent v2 | #1 |
| Harvey Legal Agent | #1 |
| Gray Swan IPI (ASR) | 0.7% |
Independent ranking: Artificial Analysis Intelligence Index
- Gemini 4 Carbon - 59
- Claude Opus 5.5 - 58
- GPT-6 Astra - 57
- GPT-6.1 Sol - 55
- Gemini 4 Argon - 50
Leaderboard checked on October 30, 2026. The figures below come from the Artificial Analysis launch article of September 30, 2026, when Argon scored 53.
- Running the index costs 1.99 USD per task on Gemini 4 Argon at the introductory price - about 60% of GPT-6 Astra (3.26 USD). At standard prices it is 3.98 USD.
- Hallucination rate 15% - the lowest among models scoring 45 or more (GPT-6 Astra: 51%).
- Argon writes long answers: on average 62k output tokens per task, versus 27k for GPT-6 Astra.
Our test
I tested Gemini 4 Argon on October 13, 2026, a few hours after general availability, on the same 23 tasks I gave Muse Wonder, Claude Opus 5.5, GPT-6 Astra, Claude Fable 5.1 and Gemini 3.8 Flash a day earlier: OpenRouter, reasoning effort low for every model, automatic grading (unit tests, tool-call checks, exact answers), one run per task, no AI judge. The set: 3 agent tasks with tools, 4 coding tasks with unit tests, 4 writing tasks in a non-English language, 11 exact-answer reasoning problems and 1 long-document task.
| Model | Score | Cost (USD) | Time (s) |
|---|---|---|---|
| Gemini 4 Argon | 99.5% | 0.22 | 134.6 |
| Muse Wonder | 99.7% | 0.64 | 164.9 |
| Claude Opus 5.5 | 99.4% | 0.75 | 222.4 |
| GPT-6 Astra | 100.0% | 1.12 | 200.1 |
| Claude Fable 5.1 | 99.4% | 1.71 | 262.8 |
| Gemini 3.8 Flash | 99.5% | 0.09 | 139.0 |
- Gemini 4 Argon scored 99.5% for 0.22 USD - the cheapest of the frontier models in this run (Muse Wonder 0.64, Claude Opus 5.5 0.75, GPT-6 Astra 1.12, Claude Fable 5.1 1.71 USD) and the fastest: 134.6 s in total.
- Its only lost point was a grading false negative in a formal letter; after a manual check the result is 100%. The real flaw in that letter: Argon added details (symptoms) that were not in the prompt.
- GPT-6 Astra was the only model at 100% by the automatic grader; Gemini 3.8 Flash also reached 99.5% for just 0.09 USD.
- Invented sources: asked for 8 real publications on each of 2 topics, Argon returned 11 of 16 real items with correct details, 3 real items with wrong details (authors or journals) and 2 invented ones; it did not invent DOIs. Claude Opus 5.5: 16 of 16, GPT-6 Astra and GPT-6.1 Sol: 15, Muse Wonder: 14, Gemini 3.8 Flash: 7. Checked in Crossref and OpenAlex.
- Tokens: one long non-English document (158,111 characters) counted as 51,482 input tokens on Argon, 87,675 on Claude and 46,929 on Muse Wonder - so the same input costs Argon roughly 40% fewer tokens than Claude.
One run per task at low reasoning effort (Google and Artificial Analysis report results at high effort): differences under 1 point are within noise. Benchmarks in the sections above are Google's own; Artificial Analysis is independent.
Articles and Analysis
- Gemini 4 Carbon is available to everyone: price, access and first test 2026-11-11
- Gemini 4 Carbon test: 50+ tasks vs Argon, Claude Fable 5.2 and GPT-6.1 Sol 2026-11-11
- GPT-6.1 Terra vs Gemini 4 Argon: Polish Test Results 2026-10-31
- IMF Warns of AI Crash as Deployments Expand 2026-10-13
- Gemini 4 Argon vs Muse Wonder: Specs and Pricing 2026-10-13
- Gemini 4 Argon Available Now: Pricing and Details 2026-10-13
- Gemini 4 Argon Context Window: Limits and Specs 2026-10-13
- Is Gemini 4 Argon Free? Access Options and Costs Explained 2026-10-13
- Gemini 4 Argon vs GPT-6.1 Sol: Which Model to Choose 2026-10-13
- Gemini 4 Argon Pricing: API Rates, Hidden Token Traps, and Costs 2026-10-13
- Gemini 4 Argon Price: What Google’s New Model Costs in 2026 2026-10-13
- AI news, week 40: September 27 - October 3, 2026 2026-10-12
Frequently Asked Questions
- What does Gemini 4 Argon cost to use?
- Until November 30, 2026, the introductory API price is $2 per 1M input tokens and $10 per 1M output tokens, with cached input at $0.10 per 1M tokens. Beginning December 1, 2026, standard pricing will rise to $4 for input, $20 for output, and $0.20 for cached input per 1M tokens. There is no free tier available in the Gemini API.
- How does Gemini 4 Argon compare with GPT-6 Astra?
- On the Artificial Analysis Intelligence Index leaderboard (checked October 30, 2026) Gemini 4 Argon scores 50 and GPT-6 Astra 57. At launch, in the Artificial Analysis article of September 30, 2026, both scored 53; Argon also had a 15% hallucination rate versus 51% for GPT-6 Astra, used on average 62k output tokens per task versus 27k, and cost 1.99 USD per task versus 3.26 USD at the introductory price.
- What benchmarks has Google reported for the model?
- According to Google, Gemini 4 Argon scored 77.9% on DeepSWE v1.1, 51.3% on AutomationBench taking 1st place, and 91.7% on LVBench long video. Google also reported a 68% score on CWE-bench v1, an 87% score on ERQA, and 1st place in the Vals Index, Vals Finance Agent v2, and Harvey Legal Agent. Google also reported a 0.7% indirect prompt injection attack success rate on Gray Swan.
- Where is Gemini 4 Argon available?
- The model is available to developers in the paid tier of the Gemini API under the model id gemini-4-argon, as well as on OpenRouter as google/gemini-4-argon. Google AI Ultra and paid Google AI Pro plans can access it in Google Antigravity (not the free plan or Google AI Plus), and it is available in the Gemini platform for business. It is also rolling out to Google AI Ultra and Pro subscribers in the Gemini app and Search AI Mode, though it may not yet appear for every Pro user.
- What are the token limits for Gemini 4 Argon?
- OpenRouter lists a context window of 1,048,576 tokens alongside a maximum completion limit of 65,536 tokens per request. Google describes a 1M unified input-output token limit via the official Gemini API and terms its output capacity industry-leading. By comparison, Google's previous generation model was limited to 64,000 output tokens.
Batch pricing and context caching
Google’s Gemini API pricing documentation lists a Batch option for Argon alongside standard paid pricing. Google describes Batch API as a 50% cost reduction, so the standard input and output rates above are not the only available rates. The same documentation specifies that Argon supports implicit context caching only. Explicit caching is not supported for this model, a distinction to account for when planning an API integration. Check the pricing page for the rates applicable to the selected mode before estimating a workload. For a broader comparison, see the AI model API prices directory.
Is Batch API cheaper for Argon?
Yes. Google lists Batch input and output rates at half the standard rates.
Can I configure explicit caching for Argon?
No. Google lists implicit caching only for Argon.
Sources for this update
API pricing and caching guidance checked on November 8, 2026.