ModelBrief · Gemini 4 Carbon · Topic Page
Last updated Nov 13, 2026
GA · NOV 10, 2026

Gemini 4 Carbon: The Topic Page

Google's strongest model, generally available since November 10, 2026: built for long-horizon agent work, complex coding, cybersecurity and scientific reasoning. The API costs the same as Argon ($2/$10 through November 30), but it spends more thinking tokens and runs slower; its Artificial Analysis index of 60 briefly topped the chart before Claude Fable 5.2 passed it two days later. This page collects the official specs, the ways to try it, pricing, and the first hands-on reviews in English and Chinese — with unverified claims clearly marked.

1M tokensContext window (1,048,576)
60AA Intelligence Index v5.0.1 (High)
$2 / $10Intro API price through Nov 30
65,536Max output tokens (streaming)
low / med / highThinking levels (medium default)
No free tierPaid API or Ultra subscription

01Key facts

What is confirmed: this is a general release, not a time-boxed preview — but its "strongest" title lasted about two days.

Announced / GA
Announced 2026-10-28 by Koray Kavukcuoglu (SVP, Google DeepMind / Google Chief AI Architect); generally available 2026-11-10
Model ID
gemini-4-carbon (stable API identifier)
Positioning
Google's strongest model: long-horizon agent workflows, complex multi-step coding, cybersecurity, scientific reasoning
Built on
Gemini 4 Argon (pre-training foundation + architecture) plus additional post-training; Google says "the main difference is how many thinking tokens each model spends per turn"
Context
1,048,576 tokens, shared by input, thinking and output
Max output
65,536 tokens (streaming); 32,768 tokens (non-streaming)
Reasoning
On by default; levels low / medium (default) / high; no minimal level (returns HTTP 400); ignores temperature, topP, topK
Input / output
Input: text, image, audio, video (plus PDF). Output: text
Knowledge cutoff
March 2026 (per the model card)
Access
Gemini API paid tier (AI Studio, Android Studio), Gemini for business, Google AI Ultra (Gemini app, AI Mode in Search, Antigravity); also third-party gateways such as OpenRouter. No free tier; Pro / Plus plans don't get it either
Price
API: $2 in / $10 out per 1M tokens through Nov 30, then $4 / $20 from Dec 1; cached input 95% off. Ultra subscription $99.99 / $199.99 per month (US)
Benchmarks
Artificial Analysis index 60 (High, v5.0.1); LMArena Text 1526 preliminary, #1; Google's official 17-test table: first in 15 (several scores self-measured by Google, see section 06C)
Weights / license
None published; proprietary hosted model, not open source
Parameters
Not disclosed

02Timeline

Thirteen days from announcement to GA; about two days at the top of the AA index.

  • 2026-10-28

    Google announces Gemini 4 Carbon on blog.google; same-day access limited to a subset of cybersecurity defenders via the Fairwind Program (the specialized cyber configuration still requires an application). The announcement claims AA index v5.0 score of 59, ranked first.

  • 2026-11-03 – 05

    LMArena Text preliminary score around 1528, provisionally #1 (~4,403 votes).

  • 2026-11-07

    Artificial Analysis index updates to v5.0.1: Carbon (High) scores 60 and takes the top spot; Opus 5.5 at 59, Astra at 57.

  • 2026-11-09

    Claude Fable 5.2 ships, scores 65 on the AA index and takes #1 from Carbon; it also leads FrontierSWE v2 at 69.9% vs Carbon's 61.2%.

  • 2026-11-10

    General availability: paid API, AI Studio, Gemini for business, Ultra subscribers (Gemini app, AI Mode, Antigravity); model card published; Antigravity launches the /swarm multi-agent preview; Cursor integrates Carbon the same day.

  • 2026-11-11 – 12

    First hands-on wave: Quantslant, Huxiu, Yupi, SeedRouter, eesel and others publish reviews in English and Chinese (see section 06).

  • 2026-11-30

    Introductory API pricing ends ($2/$10).

  • 2026-12-01

    API prices double to $4 input / $20 output per 1M tokens.

03Official specs

Official figures from the model page and model card (checked Nov 11, 2026); unpublished items are stated as such, never filled in by guesswork.

Status
GA (since 2026-11-10)
Model ID
gemini-4-carbon
Context window
1,048,576 tokens (input + thinking + output shared)
Max output
65,536 tokens (streaming); 32,768 (non-streaming)
Reasoning
On by default: low / medium (default) / high; no minimal
Input
Text, image, audio, video (plus PDF)
Output
Text
Knowledge cutoff
March 2026
API
Gemini API paid tier; no free tier
Benchmarks
AA index 60 (High, v5.0.1); LMArena 1526 (preliminary)
Model card
Published (2026-11-10)
Parameters
Not disclosed
Weights
Not published
API price
$2 / $10 through Nov 30; $4 / $20 from Dec 1 (per 1M tokens)

04How to try it: 6 routes

Developers: the paid API is fastest. Everyone else: only the Ultra subscription — Pro doesn't include Carbon.

1Gemini API paid tier (AI Studio / Android Studio)

The main developer route. Any paid-tier API key can call it as gemini-4-carbon; migrating from Argon means changing only the model string. There is no free tier — you need a billing account with prepaid credit.

2Gemini app (Ultra subscription)

Google AI Ultra subscribers get Carbon in the Gemini app, AI Mode in Search and Antigravity, with Low / Medium / High thinking levels in the model picker (High by default). Free, Plus and Pro plans don't get it.

3Gemini for business

The enterprise route: API or dropdown access, available in Antigravity and CodeMender.

4Antigravity + /swarm preview

Multi-agent orchestration preview in Google Antigravity: one prompt dispatches a coordinated team of agents; assign Carbon to architecture, coordination and auditing roles, paired with faster, lighter models and adversarial verifiers.

5Third-party gateways

On OpenRouter as google/gemini-4-carbon at the same $2/$10 price, with a :batch variant at $1/$5; 16 providers list Carbon in the models.dev catalog. Cursor integrated it on November 10.

6Scientists program / Fairwind

Scientists: selected participants in the expanded 10,000-scientists program get Argon + Carbon free for 6 months in Antigravity's highest usage tier; applications are open with weekly rolling onboarding. Cybersecurity: the specialized configuration is still gated through Fairwind Program applications.

Model: gemini-4-carbon thinking_level: low | medium (default) | high max_output_tokens: 32,768 or higher recommended (thinking and answer share the cap) HTTP timeout: 15–20 minutes recommended (long-thinking tasks)
Confirmed warning before you start: the output cap covers thinking tokens — set it too low and the model can spend everything on reasoning and return an empty answer, still billed. Quantslant measured it twice: with a 6,000-token cap, 5,808 tokens went to reasoning and the answer was empty; on another task all 16,000 tokens went to reasoning and no code came back. Google's docs give the same advice: control cost with the thinking level, not a low output cap. Also note: temperature, topP and topK are ignored; thinking_budget and thinking_level=minimal return HTTP 400.

05Pricing

Same per-token price as Argon — but not necessarily the same bill, because Carbon spends more thinking tokens per task. Prices checked Nov 11–12, 2026; the checkout page is authoritative.

API tier (per 1M tokens)Through Nov 30, 2026From Dec 1, 2026
Standard input$2.00$4.00
Standard output (thinking tokens included)$10.00$20.00
Cached input (implicit caching, 95% off)$0.10$0.20
Batch API (in / out)$1 / $5$2 / $10
Flex (in / out)$1 / $5$2 / $10
Priority (in / out)$3.60 / $18$7.20 / $36

Caching is implicit: it switches on automatically for prompts of at least 4,096 tokens. Batch / Flex / Priority rates are from a pricing-page summary (agentdeals.dev, Nov 12).

Subscription: Google AI Ultra

US pricing: $99.99/month (Ultra 5x, 5x Pro's limits) / $199.99/month (Ultra 20x). Ultra is sold in 150+ countries; Carbon reaches Ultra subscribers in "selected countries" that Google has not enumerated. Pro, Plus and free plans don't include Carbon — a tiering choice that drew complaints from paying Pro subscribers in Instagram and Facebook comments on Nov 10–11, many comparing it unfavourably with Claude Pro offering Fable 5.2.

Cost per task: cheap per token is not cheap per bill

Artificial Analysis at list price: Carbon (High) costs $7.76 per task, about 23% less than Claude Opus 5.5 (max, $10.04) and about 1.2x GPT-6 Astra (max, $6.25); Fable 5.2 (High) is $9.08 — only about a dollar more than Carbon. By thinking level, Medium at $6.59 per task is actually cheaper than Low ($7.19), because weaker settings loop more and burn more tokens. On easy work Carbon is clearly pricier: on one set of 11 small tasks, Carbon at default settings cost 3.05x Argon; but on coding-agent tasks Carbon came in at $7.74 per task against Argon's $11.73.

The pricing takeaway: through November 30 is a half-price window — the same Carbon costs twice as much per token from December 1. Test cheaply before the deadline; and remember to control cost with the thinking level (medium is the default and the best-value level), not by squeezing the output cap.

Official pages: Gemini API pricing · Google's Nov 10 GA announcement

06First hands-on reviews

All from Nov 10–12: small samples, differing methods — read each as a case study. Every card carries a verdict badge; "Read our full write-up" opens this site's in-depth rewrite, with the original source linked at the end.

English sources

Quantslant: 50+ tasks, four models head to head

Verdict: no edge on everyday work; coding is the weak spot

Source: Quantslant, Nov 11, 2026 (vs Argon, Fable 5.2, GPT-6.1 Sol; same settings, unit-test grading)

On 23 practical tasks Carbon scored 99.5% ($0.23, 264s) — tied with Argon but twice as slow; agent, spreadsheet (34/34), security review (8/8) and SOC log analysis (7/7) were all perfect. But on 6 coding tasks with 76 unit tests, Carbon passed only 63/76 while all three rivals passed 76/76. There was also a thinking-burn bug: a regex task at medium effort spent all 16,000 tokens on reasoning and returned no code.

Read our full write-up → (original reporting by this site)

SeedRouter and laozhang.ai: Carbon vs Argon

Verdict: same price, different bill — worth it only on hard jobs

Source: SeedRouter / blog.laozhang.ai, Nov 11, 2026

Same-prompt tests: both models got the answers right, but Carbon billed about 44% more output tokens (small sample); on 11 small tasks Carbon cost 3.05x Argon and took 2.75x as long. The other way round: on exact counting over a 1,500-line log, Argon answered wrong in 22 seconds while Carbon answered right in 214; on coding-agent tasks Carbon cost $7.74 per task against Argon's $11.73. Google's guidance: Argon for interactive coding, Carbon when Argon's answer misses something.

Read our full write-up → (original reporting by this site)

eesel.ai and Inferera: thinking levels and pitfalls

Verdict: medium is the best value; a low cap returns empty

Source: eesel.ai / Inferera, Nov 11–12, 2026

AA data by level: Low scores 54 at $7.19/task with a 20% made-up-answer rate; Medium scores 58 at $6.59/task with 4.2% made up (cheapest and most honest); High scores 60 at $7.76/task with 11% made up. The cost is time: about a minute before the first word, ~26 minutes per benchmark task on high. On the API side, thinking_budget, the minimal level and candidate_count>1 all return 400; a low output cap silently truncates thinking and returns empty text that is still billed; HTTP timeouts of 15–20 minutes are recommended.

Read our full write-up → (original reporting by this site)

Benchmark roundup: Google's table + independent leaderboards

Verdict: bright scores — check who ran the test

Source: Google's official Oct 28 score table; Artificial Analysis; LMArena, Nov 7–12, 2026

Google's official 17-test table: Carbon first in 15, losing only FrontierSWE 2 (61.2%, below Astra's 65.5%) and OSWorld-2.1 (81.5%, below Opus 5.5's 87.2%). But Google calculated many of the scores itself, security tests ran offline with safeguards disabled, and the comparison used the previous-generation Fable 5.1. Independent boards: AA index 60 (topped the chart Nov 7, passed by Fable 5.2's 65 on Nov 9), LMArena 1526 preliminary #1, WebDev Arena #4.

Read our full write-up → (original reporting by this site)

Chinese sources

Huxiu: nine Antigravity tests

Verdict: stable delivery, "no longer lazy"

Source: Huxiu, Nov 11, 2026 (Antigravity, High level, all one-shot)

None of the 9 tasks failed outright: a floor plan became a walkable 3D building (~1 hr), an iPhone Duo interactive 3D exhibit, hand-written cloth-physics silk scroll, a frame-by-frame remake of a 20-second promo (37 hard cuts matching the original), an 8D Chongqing night scene and more. The shortfall is finish: simple furniture, repeated materials, screen content poking out of a rounded body; the 15-second art-history piece first "cheated" by calling image generation. The author notes Google no longer claims "strongest" and instead stresses a cost per task under half of #2's, at about a quarter of Fable 5.2's price.

Read our full write-up → (original reporting by this site)

Yupi: three front-end tasks vs Fable 5.2

Verdict: disappointed — a tier below on taste and stability

Source: Tencent Cloud Developer Community, Nov 11, 2026 (Cursor, 3 parallel sub-agents, High level, same prompts as the Fable 5.2 test)

The "ancient scroll" project was interrupted 3 times by server errors and took 2 hours 16 minutes (Fable 5.2: 86 minutes, zero interruptions); the Chongqing 8D city generator shipped in 34 minutes but with only 184 boxy buildings (Fable 5.2: 1,230); the cave project had to be scrapped and restarted after nearly 2 hours, and its dog was cylinders with detached legs while the camera clipped through rock. All three projects cost $21.30, far below Fable 5.2's 500+ RMB; the author's conclusion: "don't fully trust benchmarks."

Read our full write-up → (original reporting by this site)

06BArticles: our in-depth write-ups

All 6 articles below are original rewrites and re-analysis by this site — restructured, with added conclusions and selection advice, facts and figures checked against the original reviews. Open any of them to read on-site; the original source is linked at the end of each.

Quantslant's 50+ Task Test: Carbon vs Argon, Fable 5.2 and GPT-6.1 Sol — Coding Is the Weak SpotOur deep read of Quantslant's four-model shootout: 99.5% on everyday tasks, perfect on agent and security work — but only 63 of 76 coding unit tests, where all three rivals scored 76/76; plus the thinking-burn bug that returns nothing. Huxiu's Nine Antigravity Tests: Nothing Failed — but Nothing Amazed EitherOur write-up of Huxiu's nine one-shot tests on Antigravity at High: a 3D building, an iPhone Duo exhibit, cloth physics, a frame-by-frame promo remake — stable delivery that "no longer slacks off", with finish still the weak point. Yupi's Three Front-End Tasks vs Fable 5.2: Three Server Errors, 184 Buildings Against 1,230Our write-up of Yupi's same-prompt comparison in Cursor: the ancient scroll took 2h16m, the Chongqing 8D city was a tier less detailed, the cave project had to be restarted; $21.30 for all three, and the verdict "don't fully trust benchmarks". Carbon vs Argon: Same Price, Different Bill — When Switching to Carbon PaysOur synthesis of SeedRouter's and laozhang.ai's same-prompt tests: Carbon costs 3x on easy tasks and less on hard ones; Argon answered a 1,500-line log count wrong in 22s, Carbon right in 214s. Includes a decision table. Gemini 4 Carbon Benchmarks, Fully Mapped: Google's 17-Test Table, AA 60 and LMArena 1526Our roundup of every verifiable score: Google's table (first in 15 of 17), the AA index at 60 (two days at #1), LMArena 1526, AA-Omniscience 54 — with who ran each test and what the fine print says. Thinking Levels and Pitfalls: Medium Is Cheapest and Most Honest; a Low Output Cap Returns EmptyOur guide from eesel's and Inferera's data and migration notes: score, cost and made-up-answer rates by level, why empty answers are still billed, the 400-error checklist and the 15–20 minute timeout advice.

06CBenchmarks: official table + independent boards

The opposite of Qwen4-Max: plenty of scores — the question is who ran them and against whom. First Google's official 17-test table (Oct 28, 2026), then the independent boards, then the caveats.

Test (Google's official table)CarbonArgonGPT-6 AstraOpus 5.5Fable 5.1
AA Index v5.05950575849
Vals Index v2.270.4%69.2%65.0%70.2%—
SWE-sweep9.6%4.3%9.5%3.6%2.9%
FrontierSWE 261.2%55.0%65.5%62.3%56.3%
Terminal-Bench 4.0 (AA)65.2%57.1%59.1%59.6%52.0%
Agents' Last Exam v1.177.4%69.0%73.1%71.6%—
OSWorld-2.1 (offline)81.5%74.7%80.9%87.2%76.8%
AutomationBench (Zapier)53.9%51.3%41.4%42.5%31.4%
Vending-Bench 2 (USD balance)16,48413,71815,5159,2355,422
Chartography72.5%71.6%71.0%66.3%46.2%
Blueprint-Bench 258.6%54.4%49.7%51.2%41.9%
GraphWalks 256K–1M96.2%84.2%71.8%66.8%65.0%
RiemannBench80.0%76.0%72.0%69.6%65.6%
Terminal-Bench-Science 0.1 (AA)64.8%52.9%63.3%59.0%43.3%
CWE-bench71.0%68.0%68.0%67.0%58.0%
ExploitGym54.8%42.0%42.4%34.5%—
SEC-bench Pro88.2%83.4%85.4%——

Carbon first in 15 of 17. The official table compares against the previous-generation Claude Fable 5.1, not Fable 5.2 (released Nov 9).

Independent boards and third-party measurements

MeasurementSourceCarbon's resultComparison
AA Intelligence Index v5.0.1Artificial Analysis, Nov 760.0 (High), briefly #1Fable 5.2 65 (#1 since Nov 9); Opus 5.5 59; Astra 57; Argon 50
AA index by levelArtificial Analysis / eesel, Nov 11Low 54 / Medium 58.3 / High 60Medium cheapest at $6.59/task, lowest made-up rate at 4.2%
AA-Omniscience v1.1Artificial Analysis, ~Oct 31Net score 54, #1Opus 5.5 48; accuracy 58% (Argon 48%); hallucination rate 11% (Argon 4%)
LMArena TextLMArena, Nov 6–71526 preliminary, #1 (~4,483 votes)Argon (High) 1523; Muse Wonder 1513; Opus 5.5 1505
WebDev ArenaLMArena1768 elo, #4—
CursorBench 4.0Cursor#1 (60.4% at high effort)—
Coding Agent IndexArtificial Analysis59Claude Code + Opus 5.5 at 62 — Carbon trails by 3
MKA (Multimodal Knowledge Agent)Google × AA, new benchmark#2Average cost per task under 1/6 of the top scorer's
FrontierSWE v2Snowtraces, Nov 1161.2%Fable 5.2 69.9%
Read the fine print first: in Google's official table, several scores were calculated by Google itself (SWE-sweep, Agents' Last Exam, OSWorld, GraphWalks, ExploitGym, SEC-bench Pro), security tests ran offline with safeguards disabled, and rival figures come from their own reports rather than the same run; small gaps like AA 59/58/57 or Vals 70.4/70.2 are effectively ties. Independent boards are more trustworthy but have their own conventions: the AA index rank differs depending on whether models or configurations are counted, so this page gives scores and the models ahead/behind, not a single rank number.

Full roundup: our benchmarks article →

07Comparisons

vs Gemini 4 Argon (the tier below)

DimensionGemini 4 CarbonGemini 4 Argon
StatusGA (Nov 10, 2026)GA (earlier)
API price$2 / $10 through Nov 30, then $4 / $20Identical to Carbon
Context1M1M
AA index60 (High) / 58 (Medium) / 54 (Low)50 (High)
SpeedSlow: ~26–27 min per task, ~68s to first tokenMuch faster: roughly half the time on the same tasks
SubscriptionUltra onlyPro and above
Best forThe hardest jobs: long-horizon agents, tough debugging, science reasoning, security reviewMost coding and knowledge work, interactive coding, fast loops

Google's own advice: Argon remains the right choice for most software engineering and knowledge work; keep Carbon for the hardest tasks.

Read the deep comparison: Carbon vs Argon →

vs rival flagships (Artificial Analysis index, Nov 12, 2026)

ModelAPI price $/1M (in/out)AA indexContext
Claude Fable 5.28 / 4065 (max)1M
Gemini 4 Carbon2 / 10 (through Nov 30)60 (high)1M
Claude Opus 5.5—59 (max)—
GPT-6 Astra10 / 5057 (max)1.05M
GPT-6.1 Sol2 / 1055 (max)1.05M
Qwen3.8 Max2 / 645 (Sep 27)1M
Qwen4-Max-PreviewToken Plan onlyNot listed1M
An honest note: Carbon's AA #1 lasted about two days (Nov 7–9); Fable 5.2 passed it at 65, and also leads FrontierSWE v2 (69.9% vs 61.2%). But Carbon costs a quarter of Fable 5.2's per-token price and a fifth of Astra's — its pitch is value and cost per task, not the top of the leaderboard. Note: Opus 5.5's API price and context window were not confirmed in this research round, so those cells are blank.

08What's missing (as of Nov 13)

  • Parameter count and architecture details (Google has not disclosed them)
  • Vertex AI availability (not addressed in the GA materials)
  • A country-by-country rollout schedule; the list of countries where Ultra includes Carbon
  • Exact usage limits for Ultra 5x / 20x (only prices and multiples are confirmed)
  • Independent replication of Google's self-measured benchmark scores
  • A GA date for /swarm multi-agent orchestration (currently preview)
  • Long-term reliability data (one tester hit repeated server errors on day one — a single case)
  • Pricing plans beyond the December 1 doubling

09Rumor check

No major rumor wave has formed around Carbon, but these claims are circulating or easy to get wrong.

CLARIFIED "Is Carbon's AA index 59 or 60?"

Both are real — different index versions: 59 on AA index v5.0 (at the Oct 28 announcement), 60 on v5.0.1 (after the Nov 7 update). Cite the version when you quote it.

FALSE "Carbon is #1 on the AA index"

It was, for about two days (Nov 7–9). Claude Fable 5.2 shipped on Nov 9 and leads at 65; as of Nov 12 Carbon at 60 sits behind it. Note also that different counting methods (by model vs by configuration) place Carbon anywhere from 2nd to 4th, so this page doesn't print a single rank number.

FALSE "Carbon beats Fable 5.2"

Fable 5.2 leads on the AA index (65 vs 60) and FrontierSWE v2 (69.9% vs 61.2%). Google's official comparison table used the previous-generation Fable 5.1 — it predates Fable 5.2, so citing it to claim Carbon is strongest is now out of date.

PARTLY TRUE "Carbon is just Argon with more thinking"

Google itself says Carbon is built on Argon's pre-training foundation and architecture, that "the main difference is how many thinking tokens each model spends per turn", and charges the same per-token price. But it is not the same model: the additional post-training produced real improvements — cheating attempts 3.6% → 0.4% (successful 2.4% → 0%), agentic honesty benchmark 70% → 92%, AA index 50 → 60.

FALSE "Carbon is on the AI Studio free tier / Pro plan"

The pricing page explicitly excludes Carbon from the free tier; Google AI Pro doesn't include it either — Pro still gets Argon. On the consumer side, only Ultra subscribers get Carbon.

10FAQ

When was Gemini 4 Carbon released?

Announced October 28, 2026, and generally available November 10, 2026. Between those dates it was limited to a subset of cybersecurity defenders in the Fairwind Program. See the timeline in section 02.

Is Gemini 4 Carbon free?

No. The API has no free tier; on the consumer side only Google AI Ultra subscribers ($99.99 / $199.99 per month, US) get it — Pro, Plus and free plans don't. Selected scientists get 6 months free through the scientists program.

What does the API cost — and does it go up on December 1?

Through November 30: $2 input / $10 output per 1M tokens (thinking tokens included); from December 1 the price doubles to $4 / $20. Cached input is 95% off ($0.10 → $0.20). Batch and Flex are half price. See section 05.

How is it better than Argon, and which should I use day to day?

AA index 60 vs 50, and clearly better safety and honesty metrics; but on everyday tasks it does nothing Argon can't, more slowly and at a higher bill. Google's advice: Argon for interactive coding and most knowledge work; switch to Carbon when Argon's answer misses something. See our Carbon vs Argon comparison.

Is it stronger than Claude Fable 5.2?

Not by the scores: AA index 65 vs 60 and FrontierSWE v2 69.9% vs 61.2% both favour Fable 5.2, and front-end tests also rate Fable 5.2's taste and stability higher. But Carbon costs a quarter of Fable 5.2's per-token price, and the cost per task is $7.76 vs $9.08 — only about a dollar apart.

Which thinking level should I pick — low, medium or high?

Medium, the default, is the best value: AA 58, $6.59 per task, a 4.2% made-up-answer rate (lowest of the three). High scores 60 but costs $7.76 per task with an 11% made-up rate; Low scores 54 at $7.19 per task with a 20% made-up rate. See our thinking levels and pitfalls guide.

Why did my request return an empty answer and still charge me?

The output cap covers thinking tokens: set it too low and the model spends everything on reasoning, hits the cap before writing an answer, and returns empty text with status incomplete — billed anyway. This is confirmed behaviour (Quantslant measured it; Google's docs describe it), not a random glitch. Fix: set max_output_tokens to at least 32,768 and control cost with the thinking level instead of the output cap.

How large is the context window?

1,048,576 tokens, shared by input, thinking and output. Max output is 65,536 tokens (streaming) or 32,768 (non-streaming).

Is it open source? Is it on HuggingFace?

No. No weights have been published and there is no official model on HuggingFace. It is available only through Google's API and subscription services.

Is it available on Vertex AI?

The GA materials do not address Vertex AI availability, and there was no confirmed news as of November 13. Check Google Cloud's official documentation.

Where can I follow the latest status?

Google's official blog (blog.google), the Gemini API pricing and model pages, the Artificial Analysis model page, and the source list below.

11Sources

Facts on this page come from the public sources below (checked Nov 12–13, 2026).

Official

English coverage & hands-on tests

Chinese coverage & hands-on tests