As of: 2026-11-15 00:33 CET.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/m/gemini-4-argon/

# Gemini 4 Argon

Google DeepMind, Gemini 4 family. Status: current. Type: reasoning model. Released: 2026-10-13. Training cutoff: 2026-03.

## How to call it

| Provider | Model id | Endpoint | Docs |
|---|---|---|---|
| Gemini API, generateContent (paid tier only; works with google_search grounding and audio input, tested 2026-10-25; the Interactions API also works) | `gemini-4-argon` | https://generativelanguage.googleapis.com/v1beta/models/gemini-4-argon:generateContent | https://ai.google.dev/gemini-api/docs/google-search |
| Gemini API, Interactions API (paid tier only; no Free tier; the form Google's docs samples use) | `gemini-4-argon` | https://generativelanguage.googleapis.com/v1beta/interactions | https://ai.google.dev/gemini-api/docs/models/gemini-4-argon |
| Gemini Enterprise Agent Platform (Google Cloud, location global) | `gemini-4-argon` | https://aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/global/publishers/google/models/gemini-4-argon:generateContent | https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/guides/gemini-4-argon |
| Google AI Studio (paid API tier or Google AI Ultra) |  |  | https://aistudio.google.com/prompts/new_chat?model=gemini-4-argon |
| Web app (Gemini app; Google AI Pro and Ultra subscribers) |  |  | https://gemini.google.com |
| AI Mode in Google Search (Google AI Pro and Ultra subscribers) |  |  | http://aimode.google/ |
| Google Antigravity (Google AI Ultra subscribers and eligible enterprise plans) |  |  | https://antigravity.google/ |
| GitHub Copilot (Pro+, Max, Business, Enterprise; generally available since 2026-10-13, gradual rollout, billed at provider list pricing) |  |  | https://github.blog/changelog/2026-10-13-gemini-4-argon-in-github-copilot/ |
| OpenRouter ($2 / $10 per 1M tokens; also google/gemini-4-argon:batch at $1 / $5; 1,048,576-token context; checked 2026-10-14) | `google/gemini-4-argon` |  | https://openrouter.ai/google/gemini-4-argon |
| Snowflake Cortex AI (public preview since 2026-10-13; Cortex AI Functions, cross-region inference; CoWork, CoCo and Cortex Agents "coming soon"; model id not stated) |  |  | https://www.snowflake.com/en/blog/gemini-4-argon-snowflake-cortex-ai/ |
| Google DeepMind Fairwind Program (vetted cyber defenders; no cyber guardrails) | `gemini-4-argon-cyber-permissive` |  | https://deepmind.google/fairwind-program/ |

## Pricing

- Input: $2 per 1M tokens
- Output: $10 per 1M tokens
- Cached input: $0.10 per 1M tokens
- Unit as published: per 1M tokens, paid tier only, introductory price through 2026-11-30; from 2026-12-01: $4.00 input / $20.00 output / $0.20 cached input. Output price includes thinking tokens. Batch and Flex: half price. Priority: $3.60 / $18.00 (then $7.20 / $36.00)
- Source: https://ai.google.dev/gemini-api/docs/pricing
- Last checked: 2026-10-25

## Specs

- Input: text, image, audio, video, pdf
- Output: text
- Context window: 1,048,576 tokens
- Max output: 65,536 tokens
- Open weights: no
- Licence: proprietary

## What it has not seen

- Published training cutoff: 2026-03 (March 2026).
- AI events after it on Post-Cutoff: 1,918 (439 major, 85 historic), as of 2026-11-15.
- Already missing at launch: 1,122.
- Briefings of the news since March 2026: https://postcutoff.com/briefings/cutoff-2026-03-s.md (short), https://postcutoff.com/briefings/cutoff-2026-03-m.md (medium), https://postcutoff.com/briefings/cutoff-2026-03.md (full).
- Its card, with five things it has never heard of: https://postcutoff.com/you/gemini-4-argon/
- Gemini 4 Argon is named in one of our documented cases of cutoff blindness (case 022). https://postcutoff.com/why/022-gemini-4-argon-calls-briefing-speculative-fiction/

## What stands out

- Unified 1M-token window with long decode (first): Input, thinking and output share one 1,048,576-token window. One call returns up to 65,536 tokens (32,768 for non-streaming direct HTTP on the enterprise platform); long decode continues in 64K hops up to 1M tokens minus the input, via background execution in the Interactions API or a continuation_token in generateContent. Earlier Gemini models capped output at 64K. Source: https://ai.google.dev/gemini-api/docs/long-decode
- Enterprise knowledge work: Vendor-reported: Vals Index 68.9% (#1; confirmed on the Vals AI leaderboard as of Oct 12, 2026), Harvey Legal Agent Benchmark 19.6% vs 6.7% for the next model, Vals Finance Agent v2 65.4%, Zapier AutomationBench 51.3% (#1). Source: https://deepmind.google/models/model-cards/gemini-4-argon/
- Low hallucination rate: Artificial Analysis measured a 15% hallucination rate on AA-Omniscience, the lowest of any model scoring 45+ on its Intelligence Index at the time (GPT-6 Astra 51%). Measured on the Sept 30 pre-release version. Source: https://x.com/ArtificialAnlys/status/2105392625788637299
- Frontier software engineering (mixed): Vendor-reported DeepSWE v1.1 77.9% (Opus 5.5 74.2%, GPT-6 Astra 74.1%), but it trails on FrontierSWE v2 (55.0% vs Astra 65.5%) and Terminal-Bench 4.0 (57.4% vs Opus 5.5 66.4%). Independent: Cognition FrontierCode 1.1 60.5%, level with GPT-6.1 Sol; Every found it weak at refactoring existing code. Source: https://devin.ai/blog/gemini-4-argon
- Autonomous vulnerability discovery and patching: CWE-bench v1 68% pass@1, tied first with GPT-6 Astra and Grok 4.7 (public leaderboard). A separate id, gemini-4-argon-cyber-permissive, serves Fairwind-approved defenders without cyber guardrails. Google's model card says it cannot rule out that Argon reached its cybersecurity critical capability level. Source: https://deepmind.google/models/model-cards/gemini-4-argon/
- Long-context and long-video understanding: Vendor-reported GraphWalks BFS 256K–1M 84.2% (Astra 71.8%) and LVBench 91.7%. Agentic video understanding is generally available with this model (off by default). Source: https://ai.google.dev/gemini-api/docs/latest-model
- Embodied and spatial reasoning: Vendor-reported 87% on the ERQA embodied-reasoning benchmark, which Google says is the best of any frontier model; used in a video-to-simulation-to-robot pipeline on Apptronik's Apollo humanoid. Source: https://developers.googleblog.com/gemini-4-argon-for-robotics/

## Notes

Successor: Gemini 4 Carbon (announced 2026-10-28, Fairwind cyber defenders only at first; see gemini-4-carbon), #1 on Artificial Analysis Intelligence Index v5 at 59, where Argon (high) scores 50; Argon stays the callable Gemini 4 model until Carbon is generally available. Generally available since 2026-10-13 (announced 2026-09-30 for Fairwind cyber defenders only; see 2026-09-30-gemini-4-argon). Knowledge cutoff per Google's model card: 'The knowledge cutoff for Gemini 4 is March 2026, but in some domains the model's knowledge may be less up to date.' max_output is the long-decode ceiling; a single synchronous call returns at most 65,536 tokens, and the API's models.get metadata reports outputTokenLimit 32768 (checked 2026-10-13). Thinking levels low / medium (default) / high; temperature, top_p and top_k are ignored; explicit context caching, the Live API and image or audio output are not supported. Not on the API Free tier; API and AI Studio access is not included in the Google AI Pro plan, though Pro subscribers get the model in the Gemini app and AI Mode. Frontier Safety Framework: Google cannot rule out the biological CBRN CCL, the cybersecurity CCL and the stealth/situational-awareness TCL, and says safeguards lower the risk below each. Independent: Artificial Analysis Intelligence Index 53 (Sept 30 version), Arena Text #1 at 1525 on Sept 30 (not re-checked). Safety: Google's 217-page Safety & Security Report (Oct 13, 2026, https://storage.googleapis.com/deepmind-media/gemini/gemini_4_safety_report.pdf) says it cannot rule out the unmitigated model reaching its cyber and biological critical capability levels and its stealth and situational-awareness level; external testers included UK AISI, the US NIST centre, Faculty and Apollo Research. Ids and prices re-checked against ai.google.dev on the evening of 2026-10-13: unchanged. generateContent checked 2026-10-25: models.get lists generateContent, countTokens and batchGenerateContent, and one live v1beta generateContent call with a WAV audio part plus the google_search tool answered the spoken question and returned groundingMetadata (web search queries and 2 chunks); Google's docs samples use the Interactions API, and the Google Search grounding page lists Gemini 4 Argon as supported.

Google's first Gemini 4 model and its flagship reasoning model. Model id `gemini-4-argon` (stable). Google says more Gemini 4 family
models will follow, without names or dates.

Quick start (Interactions API, from Google's docs):

```bash
curl "https://generativelanguage.googleapis.com/v1beta/interactions" \
  -H "x-goog-api-key: $GEMINI_API_KEY" -H 'Content-Type: application/json' -X POST \
  -d '{"model": "gemini-4-argon", "input": "Hello", "generation_config": {"thinking_level": "medium"}}'
```

For responses longer than 64K tokens, set `"background": true` and `max_output_tokens` above 65,536 (see the
[long decode guide](https://ai.google.dev/gemini-api/docs/long-decode)). Remove `temperature`, `top_p`, `top_k` and explicit
`cached_content` when migrating from Gemini 3.1 Pro ([migration guide](https://ai.google.dev/gemini-api/docs/latest-model)).

Benchmark highlights (Google's table in the [model card](https://deepmind.google/models/model-cards/gemini-4-argon/), with methodology in
[this PDF](https://storage.googleapis.com/deepmind-media/gemini/gemini_4_argon_model_evaluation.pdf)). Scores are for the highest
thinking setting, and rival scores are mostly the providers' own figures.

| Benchmark | Argon | GPT-6 Astra | Fable 5.1 | Opus 5.5 |
|---|---|---|---|---|
| Vals Index | 68.9 | 63.1 | 65.8 | 67.0 |
| DeepSWE v1.1 | 77.9 | 74.1 | 67.4 | 74.2 |
| FrontierSWE v2 | 55.0 | 65.5 | 56.3 | 62.3 |
| Terminal-Bench 4.0 | 57.4 | 58.2 | 57.9 | 66.4 |
| Terminal-Bench Science 0.1 | 57.6 | 68.1 | 52.6 | 63.3 |
| GraphWalks 256K–1M | 84.2 | 71.8 | 65.0 | 66.8 |
| OSWorld-2.0 (offline, partial) | 69.2 | 72.6 | — | — |
| LVBench | 91.7 | 87.5 | 79.7 | 83.7 |
| CWE-bench v1 | 68.0 | 68.0 | 58.0 | 67.0 |

Model page: https://postcutoff.com/m/gemini-4-argon/