Tech Meridian← ALL MODELS
PROMY MERIDIAN

GOOGLE / GOOGLE DEEPMIND · MODEL RELEASE TRACKER

Gemini 4 Carbon

A Gemini 4 family model (Gemini 4 Carbon). Reported to have an Artificial Analysis Intelligence Index score of 60 at the xhigh reasoning level, a token generation rate of about 57 tokens/second, and regular list pricing of $4 per million input tokens and $20 per million output tokens. It had a promotional price in effect through the end of November and an undiscounted per-task cost (xhigh) reported as $7.76.

CURRENT SNAPSHOT4/5 DIMENSIONS WITH DATA

The dimensions that change the decision.

PRICING
Promotional pricingA promotional price for Carbon was reported as being in effect through the end of November (amount not specified in the excerpt).
Regular list price per million tokens (input / output)$4 per million input tokens; $20 per million output tokens
Undiscounted per-task cost at xhigh$7.76 (undiscounted per-task figure referenced for xhigh)
CONTEXT WINDOW

Not established from the available sources.

MODALITIES
ModalityDescribed as combining frontier visual understanding with reasoning and exhaustive retrieval capabilities, indicating multimodal (text + vision) functionality.
BENCHMARKS
Artificial Analysis Intelligence Index v5.0 scoreRanks #1 on the updated Artificial Analysis Intelligence Index v5.0 with a score of 59.
Cybersecurity benchmark scoresReported scores: SEC-bench Pro (finding vulnerabilities) 88.2%, ExploitGym (exploitation) 54.8%, CWE-bench (patching) 71.0%.
Artificial Analysis Intelligence Index v5.0Ranks #1 on Artificial Analysis Intelligence Index v5.0 with a score of 59.
GDPx benchmark score10.9% on SemiAnalysis's GDPx benchmark
Rank on GDPx leaderboardRanked 2nd (behind Claude Opus 5.5) on SemiAnalysis's GDPx leaderboard
Benchmark performanceGoogle reports Carbon ranks #1 on CursorBench 4.0, Perplexity’s WANDR, and Zapier’s AutomationBench and performs strongly on the Multimodal Knowledge Agent (MKA) benchmark, providing high quality at a lower cost than comparable models in those tests.
BenchmarksReported to rank #1 on CursorBench 4.0, Perplexity’s WANDR, and Zapier’s AutomationBench; also reported to have topped the MathArena math benchmark in independent media reporting.
Relative performance (media reporting)Media reporting described Gemini 4 Carbon as a leading model in recent comparisons, noting it outperformed Mistral's Large 4 in head-to-head comparisons.
Artificial Analysis Intelligence Index score (xhigh)60
Coding Agent Index score (inside Google Antigravity)55 points
AVAILABILITY
AvailabilityAccessible via the Gemini API paid tier in Google AI Studio and Android Studio. Google AI Ultra subscribers can try Carbon in Google Antigravity, AI Studio, the Gemini app, and AI Mode in Search. Gemini for Business users can access it via the API or model dropdown, Google Antigravity, or CodeMender.

RELEASE TIMELINE

Published, source-backed release events only.

1 SOURCES · IMPORTANCE 6.0

Mistral launches Vibe Desktop app for macOS and Windows

Mistral released Vibe Desktop, a macOS and Windows app that lets its AI agent Vibe run multiple tasks in parallel, combine local Vibe CLI features with cloud sessions from Vibe Code Web, use Git worktrees to isolate parallel work, and access local files; a Linux version is planned. The company positions the release as catching up to rivals whose desktop agents (notably Anthropic’s Claude) already have deeper system-control capabilities, and Mistral’s Large 4 model still trails top competitors in head-to-head comparisons.

→
3 SOURCES · IMPORTANCE 9.0

Google DeepMind launches Gemini 4 Neon — a 1M‑token multimodal frontier model

Google DeepMind announced Gemini 4 Neon, a "frontier-level" multimodal model with a 1,000,000‑token context window. Neon is available via the Gemini API, Google AI Studio, Google Antigravity, Gemini for business, and paid Gemini app subscriptions; Google published pricing ($3 per million input tokens, $0.15 per million cached input tokens, $15 per million output tokens) and highlighted high rankings on multiple internal and external benchmarks (e.g., #3 on the Artificial Analysis Intelligence Index at xhigh with score 61, #2 on the Vals Index, and top placements on several coding, finance, legal and agent benchmarks).

→
14 SOURCES · IMPORTANCE 9.0

Google releases Gemini 4 Argon for robotics control, simulation, and policy improvement

Google released Gemini 4 Argon, a general-purpose physical-reasoning model aimed at robot control, Real2Sim scene construction, multimodal labeling of robot video, and iterative policy improvement. Google’s demos show Argon guiding a Boston Dynamics Spot and robot arms in MuJoCo twins, reconstructing scenes from teleoperation video, and synthesizing domain-randomized MuJoCo episodes (reported to raise success from 56% to 75% on one setup); Google reports Argon scores 87% on the ERQA benchmark, and makes it available to Google AI Ultra subscribers and via the paid Gemini API as gemini-4-argon, while publishing only limited per-demo failure data and relying on heavy prompt scaffolding in examples.

→

VERIFIABLE FACTS

Every value stays attached to a source and date.

CAPABILITIES · Key capabilities described by developerDEVELOPER CLAIM

Described as Google’s strongest model with state-of-the-art cybersecurity capabilities and long-horizon / sustained agentic workflows (able to run for days to pursue complex, multi-step engineering tasks); reported improvement across coding, agentic, and scientific tasks, with uplift in generating secure code and reduced misalignment and fabrication rates (per Google).

AVAILABILITY · AvailabilityDEVELOPER CLAIM

Accessible via the Gemini API paid tier in Google AI Studio and Android Studio. Google AI Ultra subscribers can try Carbon in Google Antigravity, AI Studio, the Gemini app, and AI Mode in Search. Gemini for Business users can access it via the API or model dropdown, Google Antigravity, or CodeMender.

CAPABILITIES · Primary capabilities and intended useDEVELOPER CLAIM

Described as the strongest Gemini 4 model with state-of-the-art cybersecurity and long-horizon capabilities; designed for sustained, multi-day agentic workflows to tackle complex, multi-step engineering tasks (e.g., root-cause analysis and patching subtle bugs across large codebases and low-level systems).

SAFETY · Prompt-injection resistanceDEVELOPER CLAIM

Reported to resist prompt-injection attacks better than Argon; on Gray Swan's indirect prompt-injection benchmark Google reports Carbon reduced attack success from 0.7% (Argon) to 0.0% in the tested cases, though reporting notes this does not mean prompt injection is solved.

CAPABILITIES · Capabilities and intended tasksDEVELOPER CLAIM

Described by Google as its strongest model with frontier performance in cybersecurity and long‑horizon agentic workflows across complex knowledge work, software engineering, and scientific reasoning; built for tasks that push model intelligence such as complex multi‑step coding, debugging, and open‑ended research.

BENCHMARKS · Benchmark performanceDEVELOPER CLAIM

Google reports Carbon ranks #1 on CursorBench 4.0, Perplexity’s WANDR, and Zapier’s AutomationBench and performs strongly on the Multimodal Knowledge Agent (MKA) benchmark, providing high quality at a lower cost than comparable models in those tests.

LIMITATIONS · Token usage compared to ArgonDEVELOPER CLAIM

Carbon uses more tokens than Gemini 4 Argon to reason more deeply and run extended verification loops; as a result, total task cost may be higher in practice despite the same per‑token price.

CAPABILITIES · Agent 'swarm' previewDEVELOPER CLAIM

Google is previewing a '/swarm' feature in Antigravity that turns a single prompt into a coordinated team of agents; Carbon is positioned to play deep‑reasoning roles (architecture, coordination, auditing) within such multi‑agent workflows.

WHAT CHANGED

Stored passport versions, without reconstructed history.

Passport updated32 facts
PRICING · Pricing (introductory API rate)− Introductory API pricing shared with Argon: $2 per million input tokens and $10 per million output tokens through November 30, 2026; scheduled to double to $4/$20 per million tokens on December 1, 2026 (per reporting).
CAPABILITIES · Processing time on the referenced benchmark (approx.)+ About 36 minutes (for the benchmark referenced in the excerpt)
BENCHMARKS · Coding Agent Index score (inside Google Antigravity)+ 55 points
BENCHMARKS · Artificial Analysis Intelligence Index score (xhigh)+ 60
PRICING · Undiscounted per-task cost at xhigh+ $7.76 (undiscounted per-task figure referenced for xhigh)
PRICING · Promotional pricing+ A promotional price for Carbon was reported as being in effect through the end of November (amount not specified in the excerpt).
PRICING · Regular list price per million tokens (input / output)+ $4 per million input tokens; $20 per million output tokens
CAPABILITIES · Token generation speed+ About 57 tokens per second
Passport updated26 facts
AVAILABILITY · Access for Google AI Ultra and business users− The article states Google AI Ultra subscribers can try Carbon in Google Antigravity, AI Studio, the Gemini app, and AI Mode in Search; Gemini for business users can access it via the API or model drop down, Google Antigravity, or CodeMender.
SAFETY · Alignment / safety− Described by Google as having stronger alignment and reliable follow‑through compared to prior models, intended to help with secure, production‑ready workflows.
PRICING · Pricing (introductory API rate)+ Introductory API pricing shared with Argon: $2 per million input tokens and $10 per million output tokens through November 30, 2026; scheduled to double to $4/$20 per million tokens on December 1, 2026 (per reporting).
AVAILABILITY · Availability+ Accessible via the Gemini API paid tier in Google AI Studio and Android Studio. Google AI Ultra subscribers can try Carbon in Google Antigravity, AI Studio, the Gemini app, and AI Mode in Search. Gemini for Business users can access it via the API or model dropdown, Google Antigravity, or CodeMender.
RELEASE · Broader rollout− The article states Gemini 4 Carbon 'was rolled out Carbon more broadly last week' (article published Nov 17, 2026).
CAPABILITIES · Model capabilitiesDescribed as Google’s strongest model with frontier performance in cybersecurity and long‑horizon agentic workflows across complex knowledge work, software engineering, and scientific reasoning; built for complex, multi‑step coding, debugging, and open‑ended research.→Built for tasks that push a model's intelligence; reported by Google to enable thorough exploration, deep reasoning, and extended verification loops.
PRICING · Gemini API paid tier availability− Developers can build with Carbon via the Gemini API paid tier in Google AI Studio and Android Studio (per the article).
BENCHMARKS · Relative performance (media reporting)+ Media reporting described Gemini 4 Carbon as a leading model in recent comparisons, noting it outperformed Mistral's Large 4 in head-to-head comparisons.
RELEASE · Release / rollout+ Reportedly rolled out more broadly in the week before the blog post published on 2026-11-17 (i.e., mid-November 2026).
SAFETY · Observed safety behavior in targeting tests+ In an independent evaluation of drone/surveillance targeting scenarios, Gemini 4 Carbon reportedly recognized lethal consequences and steered responses toward empty waypoints to avoid facilitating attacks.
Passport updated25 facts
AVAILABILITY · Access for Google AI Ultra and business users+ The article states Google AI Ultra subscribers can try Carbon in Google Antigravity, AI Studio, the Gemini app, and AI Mode in Search; Gemini for business users can access it via the API or model drop down, Google Antigravity, or CodeMender.
AVAILABILITY · Availability channels− Available via the API in Google AI Studio and Gemini for Business for paying developers and enterprises; rolling out to Google AI Ultra subscribers in Google Antigravity, the Gemini app, and AI Mode in Search.
RELEASE · Broader rollout+ The article states Gemini 4 Carbon 'was rolled out Carbon more broadly last week' (article published Nov 17, 2026).
CAPABILITIES · Designed for high‑intelligence tasks+ The article describes Gemini 4 Carbon as built for tasks that push the limits of a model’s intelligence.
PRICING · Gemini API paid tier availability+ Developers can build with Carbon via the Gemini API paid tier in Google AI Studio and Android Studio (per the article).
CAPABILITIES · Image-to-3D design+ The article describes a project where Carbon turned a single reference image into a 3D design (a 19th‑century optics lab with 92 instruments).
PRICING · Pricing− Introduced at the same introductory API price as Gemini 4 Argon: $2 per million input tokens and $10 per million output tokens.
CAPABILITIES · Real‑time physics simulation (example)+ The article reports builders used Carbon to simulate live physics, including an example that simulated 1,048,576 particles in real time with WebGPU.
CAPABILITIES · Reasoning and extended verification+ The article states Carbon has the ability to 'explore thoroughly, reason deeply, and run extended verification loops.'
RELEASE · Release / availability− Announced as available to developers, enterprises, scientists, and consumers (announcement published 2026-11-10).
Passport updated21 facts
SAFETY · Alignment / safety+ Described by Google as having stronger alignment and reliable follow‑through compared to prior models, intended to help with secure, production‑ready workflows.
AVAILABILITY · Availability channels+ Available via the API in Google AI Studio and Gemini for Business for paying developers and enterprises; rolling out to Google AI Ultra subscribers in Google Antigravity, the Gemini app, and AI Mode in Search.
BENCHMARKS · Benchmarks+ Reported to rank #1 on CursorBench 4.0, Perplexity’s WANDR, and Zapier’s AutomationBench; also reported to have topped the MathArena math benchmark in independent media reporting.
CAPABILITIES · CapabilitiesGoogle describes Carbon as their strongest model yet, with state-of-the-art cybersecurity defense and long-horizon engineering capabilities.→Described as Google’s strongest model with frontier performance in cybersecurity and long‑horizon agentic workflows across complex knowledge work, software engineering, and scientific reasoning; built for complex, multi‑step coding, debugging, and open‑ended research.
LIMITATIONS · Limitations / failure modes+ Independent reporting notes a common failure mode of exceeding time limits and a tendency to almost never give up on a task; an example cited a script running for more than eleven hours without checking progress.
RELEASE · General availability− Gemini 4 Carbon was released and made available to developers, enterprises, scientists, and consumers; Google made it generally available after a two‑week testing phase with selected cybersecurity defenders.
PRICING · Introductory pricing− Introduced with an introductory price equal to Gemini 4 Argon: $2 per million input tokens and $10 per million output tokens.
MODALITIES · Modality+ Described as combining frontier visual understanding with reasoning and exhaustive retrieval capabilities, indicating multimodal (text + vision) functionality.
MODALITIES · Modality− Multimodal with frontier visual understanding and reasoning across visually complex files and documents.
AVAILABILITY · Availability channels− Available via the Gemini API in Google AI Studio and Gemini for Business for paying developers and enterprises; rolling out to Google AI Ultra subscribers in Antigravity, the Gemini app, and AI Mode in Search.
PRICING · Pricing+ Introduced at the same introductory API price as Gemini 4 Argon: $2 per million input tokens and $10 per million output tokens.
RELEASE · Release / availability+ Announced as available to developers, enterprises, scientists, and consumers (announcement published 2026-11-10).
SAFETY · Safety and rollout approach− Google states it is taking a deliberate approach to Carbon's rollout and is refining and strengthening safety guardrails in tandem with partner testing.
Passport updated19 facts
RELEASE · Announcement− Gemini 4 Carbon was announced in October 2026 as the next frontier model in the Gemini 4 family.
BENCHMARKS · Benchmark performance+ Google reports Carbon ranks #1 on CursorBench 4.0, Perplexity’s WANDR, and Zapier’s AutomationBench and performs strongly on the Multimodal Knowledge Agent (MKA) benchmark, providing high quality at a lower cost than comparable models in those tests.
RELEASE · General availability+ Gemini 4 Carbon was released and made available to developers, enterprises, scientists, and consumers; Google made it generally available after a two‑week testing phase with selected cybersecurity defenders.
PRICING · Introductory pricing+ Introduced with an introductory price equal to Gemini 4 Argon: $2 per million input tokens and $10 per million output tokens.
MODALITIES · Modality+ Multimodal with frontier visual understanding and reasoning across visually complex files and documents.
MODALITIES · Modality and domain improvements− Reported improvements include scientific reasoning and multimodal understanding, along with gains in full‑stack web development and related domains (per reporting).
AVAILABILITY · Availability channels+ Available via the Gemini API in Google AI Studio and Gemini for Business for paying developers and enterprises; rolling out to Google AI Ultra subscribers in Antigravity, the Gemini app, and AI Mode in Search.
AVAILABILITY · Availability / rollout status− At the time of the announcement, Carbon was in pre-release testing with the U.S. government and a subset of trusted cyber defenders via the Fairwind Program, with partner testing being expanded ahead of a broader launch to developers, businesses, and consumers.
PRICING · Price per task (GDPx evaluation)− $16.29 per task (reported for the GDPx evaluation)
CAPABILITIES · Capabilities and intended tasks+ Described by Google as its strongest model with frontier performance in cybersecurity and long‑horizon agentic workflows across complex knowledge work, software engineering, and scientific reasoning; built for tasks that push model intelligence such as complex multi‑step coding, debugging, and open‑ended research.
CAPABILITIES · Agent 'swarm' preview+ Google is previewing a '/swarm' feature in Antigravity that turns a single prompt into a coordinated team of agents; Carbon is positioned to play deep‑reasoning roles (architecture, coordination, auditing) within such multi‑agent workflows.
LIMITATIONS · Token usage compared to Argon+ Carbon uses more tokens than Gemini 4 Argon to reason more deeply and run extended verification loops; as a result, total task cost may be higher in practice despite the same per‑token price.
Passport updated15 facts
BENCHMARKS · Rank on GDPx leaderboard+ Ranked 2nd (behind Claude Opus 5.5) on SemiAnalysis's GDPx leaderboard
BENCHMARKS · GDPx benchmark score+ 10.9% on SemiAnalysis's GDPx benchmark
PRICING · Introductory pricing− Launch price set equal to Argon: $2 per million input tokens and $10 per million output tokens; cached input tokens priced at 95% off the input token price.
PRICING · Price per task (GDPx evaluation)+ $16.29 per task (reported for the GDPx evaluation)
Passport updated13 facts
SAFETY · Alignment and fabrication− Google reports reduced misalignment and reduced fabrication (unsupported claims) rates in internal/live coding tests compared to Argon (Google states unsupported claims cut by two times vs Argon in reporting).
RELEASE · Announcement+ Gemini 4 Carbon was announced in October 2026 as the next frontier model in the Gemini 4 family.
CAPABILITIES · Capabilities+ Google describes Carbon as their strongest model yet, with state-of-the-art cybersecurity defense and long-horizon engineering capabilities.
AVAILABILITY · Initial availability− Initially available to a subset of cyber defenders and to be gradually phased to additional trusted partners via the Fairwind Program; Google is engaging in the U.S. government's voluntary pre-release model testing process.
AVAILABILITY · Availability / rollout status+ At the time of the announcement, Carbon was in pre-release testing with the U.S. government and a subset of trusted cyber defenders via the Fairwind Program, with partner testing being expanded ahead of a broader launch to developers, businesses, and consumers.
SAFETY · Safety and rollout approach+ Google states it is taking a deliberate approach to Carbon's rollout and is refining and strengthening safety guardrails in tandem with partner testing.
Passport updated11 facts
SAFETY · Alignment and fabrication+ Google reports reduced misalignment and reduced fabrication (unsupported claims) rates in internal/live coding tests compared to Argon (Google states unsupported claims cut by two times vs Argon in reporting).
RELEASE · Announcement dateAnnounced as the next-frontier Gemini 4 model on 2026-10-28.→Announced Oct 28, 2026.
BENCHMARKS · Cybersecurity benchmark scoresReported benchmark results include SEC-bench Pro: 88.2% (finding vulnerabilities), ExploitGym: 54.8% (exploitation), and CWE-bench: 71.0% (patching).→Reported scores: SEC-bench Pro (finding vulnerabilities) 88.2%, ExploitGym (exploitation) 54.8%, CWE-bench (patching) 71.0%.
CAPABILITIES · Primary capabilities and intended use+ Described as the strongest Gemini 4 model with state-of-the-art cybersecurity and long-horizon capabilities; designed for sustained, multi-day agentic workflows to tackle complex, multi-step engineering tasks (e.g., root-cause analysis and patching subtle bugs across large codebases and low-level systems).
AVAILABILITY · Initial availability+ Initially available to a subset of cyber defenders and to be gradually phased to additional trusted partners via the Fairwind Program; Google is engaging in the U.S. government's voluntary pre-release model testing process.
BENCHMARKS · Artificial Analysis Intelligence Index v5.0+ Ranks #1 on Artificial Analysis Intelligence Index v5.0 with a score of 59.
PRICING · Introductory pricing+ Launch price set equal to Argon: $2 per million input tokens and $10 per million output tokens; cached input tokens priced at 95% off the input token price.
PRICING · Introductory pricing− Introductory pricing matches Argon: $2 per million input tokens and $10 per million output tokens; cached input tokens priced at 95% off the input token price.
SAFETY · Prompt‑injection benchmark result (reported)− According to Google reporting, on Gray Swan’s indirect prompt‑injection benchmark Carbon reduced attack success rate from 0.7% (Argon) to 0.0%, reportedly blocking every tested manipulation attempt. (Reported by media citing Google.)
SAFETY · Prompt-injection resistance+ Reported to resist prompt-injection attacks better than Argon; on Gray Swan's indirect prompt-injection benchmark Google reports Carbon reduced attack success from 0.7% (Argon) to 0.0% in the tested cases, though reporting notes this does not mean prompt injection is solved.
AVAILABILITY · Rollout and initial availability− Google is taking a deliberate rollout: engaging with the U.S. government’s voluntary pre-release testing process and starting by making Carbon available to a subset of cyber defenders, then gradually granting access to additional trusted partners (via Fairwind Program) before wider availability to developers, businesses, and consumers.