Gemini 4 Argon: specs, pricing and availability

Gemini 4 Argon specs

Here are the core specifications for gemini-4-argon, which has a March 2026 knowledge cutoff and takes text, image, video, audio, and PDF inputs.

Google describes Gemini 4 Argon as a flagship reasoning model built for complex software engineering, deep mathematical and scientific problem solving, autonomous multi-step agents, and long-decode generation up to 1M tokens.

According to the model card, it uses a sparse mixture-of-experts transformer architecture and was trained on TPUs using JAX and ML Pathways. It isn't a fine-tune or modification of an earlier model.

Gemini 4 Argon pricing

The Gemini 4 Argon API has no free tier, bills for every token the model processes including thinking tokens, and doubles in price on December 1, 2026, as the table below shows (see Gemini API pricing for every model's prices).

• Implicit caching turns on automatically for prompts of 4,096 tokens or more and cuts cached input costs by 95%, while explicit caching isn't supported.
• Grounding with Google Search includes 5,000 free search requests per month shared across all Gemini 3 and newer models, then costs $14 per 1,000 requests with a daily cap of 1,500 requests.

Gemini 4 Argon benchmarks

The table below shows Google's own numbers from the Gemini 4 Argon model card, comparing it against GPT-6 Astra and Claude Opus 5.5.

• Argon leads the group on agent, finance, legal, and long-context evaluations, including AutomationBench, Vals Finance Agent v2, and GraphWalks.
• It trails on FrontierSWE v2, where GPT-6 Astra scores 65.5, and on Terminal-bench 4.0, where Claude Opus 5.5 hits 66.4.

Where you can use Gemini 4 Argon

The table below covers where Gemini 4 Argon is available and what each product requires (for a breakdown of the consumer tiers, see Google AI Pro vs Ultra).

Google also announced Gemini 4 Carbon on October 28, 2026. Carbon is a separate model restricted to cyber defenders through Google's Fairwind Program, and Google says it will launch at Argon's introductory price.

API behavior to know

Gemini 4 Argon supports low, medium, and high thinking levels, and defaults to medium. The API ignores temperature, top_p, and top_k, so you can take those out of your payloads, and if you want to cut token usage, lower the thinking level instead of tightening the output token limit.

Because max_output_tokens covers both thinking tokens and the final answer, any existing app with a limit under 32K should raise it to at least 32768 so the model doesn't run out of budget mid-thought. For heavy reasoning or coding tasks that need long outputs, set background to true and set max_output_tokens to 131072 or 262144 (or up to 1048576 minus your input tokens); if you set background to true and leave max_output_tokens unset, the server stops at 64K tokens. Long responses generate in hops of up to 64K tokens, and Google recommends setting your HTTP timeouts to at least 15 minutes.

On the tool side, the model supports code execution, computer use (preview), file search, function calling, Grounding with Google Maps, Grounding with Google Search, structured outputs, URL context, and agentic video understanding. It doesn't support audio generation, image generation, or the Live API.

Gemini 4 Argon through a Google AI plan

If you already pay for a Google AI Pro or Google AI Ultra subscription, Plan2API lets you call Gemini 4 Argon through a Gemini subscription API without per-token billing. Once you link your Google account at plan2api.com/connect, Plan2API exposes the model through Antigravity as gemini/gemini-4-argon-high (the google/ and antigravity/ prefixes work too) on an OpenAI-compatible Chat Completions endpoint (POST /v1/chat/completions) at https://api.plan2api.com/v1, with the full 1,048,576-token context window and support for text, image, audio, and video inputs.

Requests run against your Google AI plan's own usage limits rather than a per-token meter, and if you connect more than one account, a request that hits a limit moves to the next account before any output is sent. When every connected account is out of quota, the API returns HTTP 429 with the error code subscription_quota_exhausted and a Retry-After header in seconds for the earliest reset. The curl example below sends a first request.

curl https://api.plan2api.com/v1/chat/completions \
  -H "Authorization: Bearer $PLAN2API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "gemini/gemini-4-argon-high", "messages": [{"role": "user", "content": "Review this function for bugs"}]}'