Skip to main content
AI ModelsEvidence

Model releases, capability-per-dollar, training compute & safety posture

The frontier is a four-lab race (Anthropic, OpenAI, Google, Meta) measured four ways: release cadence by quarter, capability-per-dollar (a top model scores ~62 on the AA Intelligence Index at ~$25/M output tokens), training compute rising off a ~3.1e23-FLOP GPT-3 baseline (2020), and a safety-posture matrix.

The pattern: capability keeps climbing while the price of a given capability level keeps falling, the two moving in opposite directions is the whole story.

The Frontier AI Model Race (2026)

Data date
June 2026
Source class
Official body or company filing
Data points
21
Sources
8
Value basis
21 actual

Caveat: FLOP figures are third-party estimates (Epoch AI), not disclosed by labs.

This page can answer

Four series on the 2026 frontier-model race: quarterly releases by lab, a capability-vs-cost scatter of leading models, an estimated training-compute (FLOP) trend, and an editorial safety-posture matrix across risk domains.

It cannot answer

  • Exact, lab-disclosed training compute (FLOP figures are third-party estimates).
  • Real-world safety outcomes (the matrix is a posture summary on an editorial scale).
  • Market share or revenue by lab (not in this index).
Jump to chaptersSource ledger
Evidence chapters

Chapter 01 · Question answered

Count of frontier releases per quarter by lab.

Model release cadence by lab

Releases by quarter

Releases by quarter

Frontier-model releases per quarter, by lab, 2024 → 2026.

View data & sources →

Releases by quarter

Rendering the interactive chart…

Expand data table
Download CSV
Model release cadence by lab: data table
PeriodSeriesValueValue basisSource
2024 Q2Meta0ActualAnthropic Model Cards & Release Notes
2024 Q2Google0ActualAnthropic Model Cards & Release Notes
2024 Q2Openai1ActualAnthropic Model Cards & Release Notes
2024 Q2Anthropic1ActualAnthropic Model Cards & Release Notes
2024 Q3Meta1ActualOpenAI Research Blog
2024 Q3Google0ActualOpenAI Research Blog
2024 Q3Openai1ActualOpenAI Research Blog
2024 Q3Anthropic0ActualOpenAI Research Blog
2024 Q4Meta0ActualGoogle DeepMind Research
2024 Q4Google1ActualGoogle DeepMind Research
2024 Q4Openai1ActualGoogle DeepMind Research
2024 Q4Anthropic0ActualGoogle DeepMind Research
2025 Q1Meta0ActualAnthropic Model Cards & Release Notes
2025 Q1Google0ActualAnthropic Model Cards & Release Notes
2025 Q1Openai0ActualAnthropic Model Cards & Release Notes
2025 Q1Anthropic1ActualAnthropic Model Cards & Release Notes
2025 Q2Meta1ActualMeta AI: Llama release notes
2025 Q2Google0ActualMeta AI: Llama release notes
2025 Q2Openai0ActualMeta AI: Llama release notes
2025 Q2Anthropic1ActualMeta AI: Llama release notes
2025 Q3Meta0ActualOpenAI Research Blog
2025 Q3Google0ActualOpenAI Research Blog
2025 Q3Openai1ActualOpenAI Research Blog
2025 Q3Anthropic0ActualOpenAI Research Blog
2025 Q4Meta0ActualGoogle DeepMind Research
2025 Q4Google1ActualGoogle DeepMind Research
2025 Q4Openai0ActualGoogle DeepMind Research
2025 Q4Anthropic0ActualGoogle DeepMind Research
2026 Q1Meta0ActualAnthropic Model Cards & Release Notes
2026 Q1Google0ActualAnthropic Model Cards & Release Notes
2026 Q1Openai0ActualAnthropic Model Cards & Release Notes
2026 Q1Anthropic1ActualAnthropic Model Cards & Release Notes
2026 Q2Meta0ActualAnthropic Model Cards & Release Notes
2026 Q2Google1ActualAnthropic Model Cards & Release Notes
2026 Q2Openai1ActualAnthropic Model Cards & Release Notes
2026 Q2Anthropic2ActualAnthropic Model Cards & Release Notes

Model release cadence by lab: full data table

SOURCEAnthropic Model Cards & Release Notes · OpenAI Research Blog · Google DeepMind Research · +1 more
VALUE BASISActual
LAST VERIFIEDSep 10, 2026
How this is calculated

Chapter 02 · Question answered

Benchmark score vs output-token price for leading models.

Capability vs cost

Capability vs cost

Capability vs cost

Artificial Analysis Intelligence Index vs output $/M tokens (June 2026 snapshot).

View data & sources →

Capability vs cost

Rendering the interactive chart…

Expand data table
Download CSV
Capability vs cost: data table
DimensionSeriesValueValue basisSource
Claude Opus 4.8Benchmark score62ActualArtificial Analysis: model intelligence & pricing
Claude Opus 4.8Cost per m tokens25ActualArtificial Analysis: model intelligence & pricing
GPT-5.5Benchmark score60ActualArtificial Analysis: model intelligence & pricing
GPT-5.5Cost per m tokens12ActualArtificial Analysis: model intelligence & pricing
Gemini 3.1 ProBenchmark score57ActualArtificial Analysis: model intelligence & pricing
Gemini 3.1 ProCost per m tokens12ActualArtificial Analysis: model intelligence & pricing
Claude Sonnet 4.6Benchmark score51ActualArtificial Analysis: model intelligence & pricing
Claude Sonnet 4.6Cost per m tokens15ActualArtificial Analysis: model intelligence & pricing

Capability vs cost: full data table

SOURCEArtificial Analysis: model intelligence & pricing
VALUE BASISActual
LAST VERIFIEDSep 10, 2026
How this is calculated

Chapter 03 · Question answered

Estimated training FLOP over time.

Training compute growth

Training compute

Training compute

Estimated training compute of frontier models by year (Epoch AI).

View data & sources →

Training compute

Rendering the interactive chart…

Expand data table
Download CSV
Training compute growth: data table
PeriodSeriesValueValue basisSource
2020Flops estimate3.1e+23ActualEpoch AI: compute & training trends
2022Flops estimate2.5e+24ActualEpoch AI: compute & training trends
2023Flops estimate2e+25ActualEpoch AI: compute & training trends
2024Flops estimate5e+25ActualEpoch AI: compute & training trends
2025Flops estimate5e+26ActualEpoch AI: compute & training trends

Training compute growth: full data table

SOURCEEpoch AI: compute & training trends
VALUE BASISActual
LAST VERIFIEDSep 10, 2026

Caveat: Flop figures are third-party estimates (Epoch AI), not disclosed by labs.

How this is calculated

Chapter 04 · Question answered

Per-model risk-domain posture (blocked / gated / allowed).

Safety-gated capabilities matrix

Safety posture matrix

Safety posture matrix

Editorial encoding of published model-safety policies (allowed / gated / blocked).

View data & sources →

Safety posture matrix

Rendering the interactive chart…

Expand data table
Download CSV
Safety-gated capabilities matrix: data table
DimensionSeriesValueValue basisSource
Claude Fable 5Bio1ActualAnthropic Model Cards & Release Notes
Claude Fable 5Chem1ActualAnthropic Model Cards & Release Notes
Claude Fable 5Cyber1ActualAnthropic Model Cards & Release Notes
Claude Fable 5Legal3ActualAnthropic Model Cards & Release Notes
Claude Fable 5Medical2ActualAnthropic Model Cards & Release Notes
Claude Opus 4.8Bio2ActualAnthropic Model Cards & Release Notes
Claude Opus 4.8Chem2ActualAnthropic Model Cards & Release Notes
Claude Opus 4.8Cyber2ActualAnthropic Model Cards & Release Notes
Claude Opus 4.8Legal3ActualAnthropic Model Cards & Release Notes
Claude Opus 4.8Medical2ActualAnthropic Model Cards & Release Notes
Frontier baseline (typical)Bio2ActualStanford HAI: AI Index Report
Frontier baseline (typical)Chem2ActualStanford HAI: AI Index Report
Frontier baseline (typical)Cyber2ActualStanford HAI: AI Index Report
Frontier baseline (typical)Legal3ActualStanford HAI: AI Index Report
Frontier baseline (typical)Medical2ActualStanford HAI: AI Index Report

Safety-gated capabilities matrix: full data table

SOURCEAnthropic Model Cards & Release Notes · Stanford HAI: AI Index Report
VALUE BASISActual
LAST VERIFIEDSep 10, 2026

Caveat: Editorial 1=blocked / 2=gated / 3=allowed scale, a posture summary, not a measured benchmark.

How this is calculated
All data for this index

Data table

Download CSV
The Frontier AI Model Race (2026): full data table
MetaGoogleOpenaiSeriesQuarterAnthropicSourceValue basisModelBenchmark scoreCost per m tokensYearFlops estimateBioChemCyberLegalMedical
001releases2024 Q21Anthropic Model Cards & Release NotesActualNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot published
101releases2024 Q30OpenAI Research BlogActualNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot published
011releases2024 Q40Google DeepMind ResearchActualNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot published
000releases2025 Q11Anthropic Model Cards & Release NotesActualNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot published
100releases2025 Q21Meta AI: Llama release notesActualNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot published
001releases2025 Q30OpenAI Research BlogActualNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot published
010releases2025 Q40Google DeepMind ResearchActualNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot published
000releases2026 Q11Anthropic Model Cards & Release NotesActualNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot published
011releases2026 Q22Anthropic Model Cards & Release NotesActualNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot published
Not publishedNot publishedNot publishedcapability_costNot publishedNot publishedArtificial Analysis: model intelligence & pricingActualClaude Opus 4.86225Not publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot published
Not publishedNot publishedNot publishedcapability_costNot publishedNot publishedArtificial Analysis: model intelligence & pricingActualGPT-5.56012Not publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot published
Not publishedNot publishedNot publishedcapability_costNot publishedNot publishedArtificial Analysis: model intelligence & pricingActualGemini 3.1 Pro5712Not publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot published
Not publishedNot publishedNot publishedcapability_costNot publishedNot publishedArtificial Analysis: model intelligence & pricingActualClaude Sonnet 4.65115Not publishedNot publishedNot publishedNot publishedNot publishedNot publishedNot published
Not publishedNot publishedNot publishedtraining_computeNot publishedNot publishedEpoch AI: compute & training trendsActualNot publishedNot publishedNot published20203.1e+23Not publishedNot publishedNot publishedNot publishedNot published
Not publishedNot publishedNot publishedtraining_computeNot publishedNot publishedEpoch AI: compute & training trendsActualNot publishedNot publishedNot published20222.5e+24Not publishedNot publishedNot publishedNot publishedNot published
Not publishedNot publishedNot publishedtraining_computeNot publishedNot publishedEpoch AI: compute & training trendsActualNot publishedNot publishedNot published20232e+25Not publishedNot publishedNot publishedNot publishedNot published
Not publishedNot publishedNot publishedtraining_computeNot publishedNot publishedEpoch AI: compute & training trendsActualNot publishedNot publishedNot published20245e+25Not publishedNot publishedNot publishedNot publishedNot published
Not publishedNot publishedNot publishedtraining_computeNot publishedNot publishedEpoch AI: compute & training trendsActualNot publishedNot publishedNot published20255e+26Not publishedNot publishedNot publishedNot publishedNot published
Not publishedNot publishedNot publishedsafety_matrixNot publishedNot publishedAnthropic Model Cards & Release NotesActualClaude Fable 5Not publishedNot publishedNot publishedNot published11132
Not publishedNot publishedNot publishedsafety_matrixNot publishedNot publishedAnthropic Model Cards & Release NotesActualClaude Opus 4.8Not publishedNot publishedNot publishedNot published22232
Not publishedNot publishedNot publishedsafety_matrixNot publishedNot publishedStanford HAI: AI Index ReportActualFrontier baseline (typical)Not publishedNot publishedNot publishedNot published22232

Method

How this evidence was built

What it measures

Source-backed values are seeded for four of the five charts: the release cadence by lab (2024 → 2026, from each lab’s release notes), the capability-vs-cost scatter (Artificial Analysis Intelligence Index vs output $/M tokens, June 2026 snapshot), training-compute growth by year (Epoch AI, corroborated by Stanford hai), and a safety-gated capabilities matrix.

How it is calculated

  1. This is an interpretation of each model’s published safety policy, not a measured benchmark.
  2. Placeholder: the per-lab benchmark-trajectory chart is left unseeded, a consistent historical per-quarter, per-lab benchmark series was not sourceable without mixing incompatible benchmarks.

Key context

  • Every numeric point carries a source citation and a basis label.
  • Sources are each lab’s own model cards/release notes, Artificial Analysis, Epoch AI, and Stanford hai: cross-checked against public release timelines.
  • Editorial encoding: the safety-gated matrix scores each model × domain as 3 = allowed / 2 = gated / 1 = blocked.

Key numbers

  • Four labs trade the lead quarter to quarter: releases are clustered, not steady (anthropic-news / openai-research / google-deepmind / meta-ai).
  • A leading model tops the Artificial Analysis Intelligence Index at ~62 while costing ~$25/M output tokens, capability up, price-per-capability down (artificial-analysis).
  • Training compute scales off a ~3.1e23-flop gpt-3 (2020) baseline (epoch-ai, corroborated by Stanford hai).
  • The safety matrix shows hard limits blocking high-risk cyber/bio/chem while gating legal/medical use, an editorial, not measured, scale (anthropic-news).

Caveats

Caveat: Flop figures are third-party estimates (Epoch AI), not disclosed by labs.

Caveat: Editorial 1=blocked / 2=gated / 3=allowed scale, a posture summary, not a measured benchmark.

Do not use this to

  • Exact, lab-disclosed training compute (flop figures are third-party estimates).
  • Real-world safety outcomes (the matrix is a posture summary on an editorial scale).
  • Market share or revenue by lab (not in this index).

Scope

June 2026

Sources

Last verified 10 Sept 2026

Methods that use this evidence

Use one of these copy-paste methods when you want an AI to work from this evidence.

How relevant was this information?