Current modelsstable
GPT-6 Luna
Luna targets focused requests at large volume, including workloads sensitive to token cost.
Observation participant · checking evidence
Not enough timing points yet
THE MODEL PERFORMANCE BROWSER
Collecting baselineAI model performance
Explore current AI models, regional coverage and recent latency evidence.
OpenAIChatGPT
—latest health TTFT7 catalog3 current1 monitoredChecking regions
AnthropicClaude
—latest health TTFT8 catalog4 current1 monitoredChecking regions
GoogleGemini
—latest health TTFT11 catalog2 current1 monitoredChecking regions
26 verified catalog entries · 3 observation participants
OpenAI / ChatGPT
3 current · 1 monitored · 7 catalog models
—latest health TTFT
GPT models for reasoning, coding and everyday work. API observations are separate from consumer app performance.
Current models 3
Current modelsstable · active
GPT-6 Astra
OpenAI positions Astra for demanding reasoning, software development, research and document work.
Catalog only
Current modelsstable · active
GPT-6.1 Sol
Sol is the current OpenAI option for complex work with a lower token price than Astra.
Catalog only
Previous generation 4
Previous generationstable · active
GPT-6 Sol
The earlier Sol model supports programming and agent workflows. OpenAI directs users to GPT-6.1 Sol for the newer version.
Catalog only
Previous generationstable · active
GPT-5.6 Sol
This is the flagship tier within the earlier GPT-5.6 family, aimed at complex professional tasks.
Catalog only
Previous generationstable · active
GPT-5.6 Terra
Terra occupies the cost-balanced tier of the GPT-5.6 family; OpenAI relates its role to earlier mini tiers.
Catalog only
Previous generationstable · active
GPT-5.6 Luna
The earlier Luna model targets inexpensive, high-volume work; OpenAI relates its role to earlier nano tiers.
Catalog only
Anthropic / Claude
4 current · 1 monitored · 8 catalog models
—latest health TTFT
Claude models for language, reasoning and extended agent workflows. API observations are separate from consumer app performance.
Current models 4
Current modelsstable
Claude Haiku 4.5
Haiku is the smaller current Claude option. It uses manually enabled extended thinking, rather than the adaptive mechanism in newer Claude models.
Observation participant · checking evidence
Not enough timing points yet
Current modelsstable · active
Claude Fable 5.1
Fable is intended for demanding reasoning and agent work that spans many steps, including coding, research and document tasks.
Catalog only
Current modelsstable · active
Claude Opus 5.5
Opus supports extended coding-agent sessions and knowledge work. Its thinking behavior differs from the earlier Opus 5 API.
Catalog only
Current modelsstable · active
Claude Sonnet 5.5
Sonnet is positioned between capability and responsiveness. Its API changes include thinking behavior and tool-use compatibility.
Catalog only
Previous generation 3
Previous generationstable · active
Claude Fable 5
An earlier Fable release remains listed as active in Anthropic’s lifecycle table. See the current family above for its successor.
Catalog only
Previous generationstable · active
Claude Opus 5
An earlier Opus release remains listed as active in Anthropic’s lifecycle table. See the current family above for its successor.
Catalog only
Previous generationstable · active
Claude Sonnet 5
An earlier Sonnet release remains listed as active in Anthropic’s lifecycle table. See the current family above for its successor.
Catalog only
Deprecated models 1
Deprecated modelsstable · deprecated
Claude Sonnet 4.5
Anthropic deprecated this Sonnet snapshot on September 30, 2026 and names Sonnet 5.5 as its replacement.
Catalog only
Google / Gemini
2 current · 1 monitored · 11 catalog models
—latest health TTFT
Gemini models for multimodal applications and responsive AI experiences. API observations are separate from consumer app performance.
Current models 2
Current modelsstable
Gemini 3.5 Flash-Lite
Flash-Lite is intended for inexpensive, high-throughput execution, including document parsing and subagent requests.
Observation participant · checking evidence
Not enough timing points yet
Current modelsstable · active
Gemini 3.8 Flash
Google’s latest stable Flash targets software engineering agents and extended enterprise workflows.
Catalog only
Previous generation 7
Previous generationstable · active
Gemini 3.7 Flash
Google labels 3.7 Flash as an earlier generation for coding, agent work and multi-step execution.
Catalog only
Previous generationstable · active
Gemini 3.6 Flash
This earlier Flash release supports code generation, iterative agent requests and spatial reasoning.
Catalog only
Previous generationstable · active
Gemini 3.5 Flash
The model overview places 3.5 Flash in the legacy Flash lineup. Its model page describes subagent use and multi-step tasks.
Catalog only
Previous generationstable · active
Gemini 3.1 Flash-Lite
This earlier Flash-Lite serves lightweight, frequent requests and data extraction with multimodal inputs.
Catalog only
Previous generationstable · limited
Gemini 2.5 Pro
The 2.5 Pro documentation describes reasoning and coding use for complex tasks.
Catalog only
Previous generationstable · limited
Gemini 2.5 Flash
The 2.5 Flash listing describes reasoning tasks with high request volume.
Catalog only
Previous generationstable · limited
Gemini 2.5 Flash-Lite
Flash-Lite is the smaller multimodal option in the 2.5 family.
Catalog only
Preview models 2
Preview modelspreview · active
Gemini 3.1 Pro Preview
This preview focuses on improved reasoning, software engineering and agent workflows using tools.
Catalog only
Preview modelspreview · active
Gemini 3 Flash Preview
This Flash preview combines multimodal understanding with reasoning and agent-oriented coding.
Catalog only
Model comparison
Individual observations · no speed ranking| Model | Provider | Lifecycle | Monitoring | Health TTFT | Speed TTFT | Regions | Last measured |
|---|---|---|---|---|---|---|---|
| GPT-6 Astra | OpenAI | current | Catalog only | — | — | 0 | Not measured |
| GPT-6.1 Sol | OpenAI | current | Catalog only | — | — | 0 | Not measured |
| GPT-6 Luna | OpenAI | current | Checking evidence | — | — | — | Awaiting evidence |
| GPT-6 Sol | OpenAI | previous | Catalog only | — | — | 0 | Not measured |
| GPT-5.6 Sol | OpenAI | previous | Catalog only | — | — | 0 | Not measured |
| GPT-5.6 Terra | OpenAI | previous | Catalog only | — | — | 0 | Not measured |
| GPT-5.6 Luna | OpenAI | previous | Catalog only | — | — | 0 | Not measured |
| Claude Fable 5.1 | Anthropic | current | Catalog only | — | — | 0 | Not measured |
| Claude Opus 5.5 | Anthropic | current | Catalog only | — | — | 0 | Not measured |
| Claude Sonnet 5.5 | Anthropic | current | Catalog only | — | — | 0 | Not measured |
| Claude Haiku 4.5 | Anthropic | current | Checking evidence | — | — | — | Awaiting evidence |
| Claude Fable 5 | Anthropic | previous | Catalog only | — | — | 0 | Not measured |
| Claude Opus 5 | Anthropic | previous | Catalog only | — | — | 0 | Not measured |
| Claude Sonnet 5 | Anthropic | previous | Catalog only | — | — | 0 | Not measured |
| Claude Sonnet 4.5 | Anthropic | deprecated | Catalog only | — | — | 0 | Not measured |
| Gemini 3.8 Flash | Google | current | Catalog only | — | — | 0 | Not measured |
| Gemini 3.5 Flash-Lite | Google | current | Checking evidence | — | — | — | Awaiting evidence |
| Gemini 3.7 Flash | Google | previous | Catalog only | — | — | 0 | Not measured |
| Gemini 3.6 Flash | Google | previous | Catalog only | — | — | 0 | Not measured |
| Gemini 3.5 Flash | Google | previous | Catalog only | — | — | 0 | Not measured |
| Gemini 3.1 Flash-Lite | Google | previous | Catalog only | — | — | 0 | Not measured |
| Gemini 2.5 Pro | Google | previous | Catalog only | — | — | 0 | Not measured |
| Gemini 2.5 Flash | Google | previous | Catalog only | — | — | 0 | Not measured |
| Gemini 2.5 Flash-Lite | Google | previous | Catalog only | — | — | 0 | Not measured |
| Gemini 3.1 Pro Preview | Google | preview | Catalog only | — | — | 0 | Not measured |
| Gemini 3 Flash Preview | Google | preview | Catalog only | — | — | 0 | Not measured |
How to read this page
Health TTFTTime to first readable text in a short availability check.
Speed TTFTA separate, longer request profile. Keep it separate from health.
FreshVerified observation within its freshness window. Green does not mean fast.
HistoricalPrevious evidence with its measurement time and region.
Catalog onlyOfficial model information only. Monitoring has not started.
Previous generationEarlier provider lineup, preserved separately from current models.