Gemini 4 Argon GA on Oct 13: Model ID, Pricing and Free-Key Errors
Gemini 4 Argon reached general availability (GA) on October 13, 2026. Its model ID is gemini-4-argon. To use it you must call it on a paid tier of the Gemini API. The free tier is not supported, so calls made with a free-tier key will fail. Below are the timeline, specs, pricing and troubleshooting steps, based on the official pages as read on October 17, 2026.
How was Gemini 4 Argon rolled out?
- September 30: Google announced Gemini 4 Argon. At that point it was offered only to trusted cybersecurity defenders through the Fairwind program, with a stated plan to widen access in stages (official announcement).
- October 13: the Gemini API changelog recorded
gemini-4-argonas GA (API changelog). The same day, Google's developer blog said you can start building on the Gemini API paid tier, in Google AI Studio and in Android Studio. Google AI Ultra subscribers can try it in AI Studio and Antigravity (developer post). - Also that day, the Gemini app opened to Google AI Pro and Ultra subscribers (Gemini app announcement). That page does not say whether free users can access it, so we make no claim either way.
What specs does the official model page list?
The following comes from the model page (marked as updated in October 2026):
| Item | Details |
|---|---|
| Model ID | gemini-4-argon |
| Context window | 1,048,576 tokens (input and output combined) |
| Max output | 65,536 by default for synchronous calls; up to 1,048,576 with long decode enabled |
| Input | Text, images, video, audio, PDF |
| Output | Text only |
| Supported | Thinking (low / medium / high), function calling, structured output, Search grounding, URL context, code execution, File search, Google Maps grounding, computer use (Preview) |
| Caching | Implicit caching only; explicit caching is not supported |
| Call modes | Standard, Batch, Flex, Priority |
| Not supported | Live API, image generation, audio generation |
If your app depends on the Live API or explicit caching, you can't move to this model directly yet.
Is Gemini 4 Argon free? What does it cost?
It is not free. The pricing page states it is not available on the Free tier. Prices per million tokens (USD) are below. The first number applies through November 30, 2026; the second applies from December 1.
| Tier | Input | Output (including thinking) |
|---|---|---|
| Standard | 2 → 4 | 10 → 20 |
| Batch / Flex | 1 → 2 | 5 → 10 |
| Priority | 3.6 → 7.2 | 18 → 36 |
A few things to note:
- Prices double on December 1. For long-term budgets, use the second number.
- Output pricing includes thinking. Setting thinking to high will raise your bill; actual token consumption depends on real usage.
- The pricing page does not list a separate price for long decode. Output can reach about 1 million tokens, so at the December price of $20 per million tokens, a single response that uses the full limit would cost about $20. This is our arithmetic from the published unit price, not a number Google gives. Set an output cap before going to production.
- Official benchmark results (for example, DeepSWE v1.1 at 77.9% in the announcement) are Google's claims as of September 30. Whether the model suits your workload is something to test on your own samples.
How do you use it? A minimal call
Step one: in Google AI Studio, attach a billing account to your project so it moves to a paid tier. Check the official docs for the exact entry point and tier names. Step two: send a request with that project's key:
curl "https://generativelanguage.googleapis.com/v1beta/models/gemini-4-argon:generateContent" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H "Content-Type: application/json" \
-d '{"contents":[{"parts":[{"text":"Explain the TCP three-way handshake in one sentence"}]}]}'
For the parameter names that control thinking level and long decode, refer to the model page, the official developer docs and the SDK docs; we don't guess them here. Before using it, make sure your region and account comply with Google's terms of service and local regulations.
Is a failed call a key problem or a network problem?
Check in this order:
- Look at the response. If you already got an HTTP 4xx with an error message, the request reached Google. It is most likely a permission, quota or model-name problem, not the network.
- Confirm the key's project has a billing account attached. A free-tier key calling
gemini-4-argonwill fail. - Verify the model ID is spelled exactly
gemini-4-argon. - Only if there is no response at all, a timeout, or a TLS handshake failure is it a network problem. Open the cdnping homepage and test the Gemini API's first-connection and round-trip times. Then check the Cloudflare edge probes of the official endpoint at /en/status/gemini to see whether the problem is on Google's side or only on yours.
- If latency is high, separate connection setup, gateway round trip and time to first token. See AI API latency breakdown.
Latency means different things on different tiers. Don't mistake a model that is slow to think for a slow network.
How do you keep tracking changes after launch?
In a new model's first few weeks, prices, limits and docs often change. You can use /en/changes to follow additions and removals on the official pricing pages and changelogs; for the approach, see how to find out right away when LLM API prices change. If you use models from several vendors, multi-model routing can split traffic by task difficulty and reduce the share going to expensive models. When a service misbehaves, how to read official status pages helps you judge incident severity and real impact.
