Confirm Action

Are you sure you want to proceed?

29 dated · 29 model families in the current roster

AI model releases over time

A release-date view of the model families in Spring Prompt's current business-skills evaluation roster, including cadence by model line and provider.

Release history, not a leaderboard. This page intentionally contains no unpublished benchmark score, rank, verdict, or claim that newer means better.

29
Models dated
GPT-5.6 Luna
Newest · 1 month ago
GPT-5 Mini
Oldest · 1.0 years ago

Release history by provider

Every model, by age

Model Released ▾
GPT-5.6 LunaOpenAI 1 month ago(2026-07-09)
GPT-5.6 SolOpenAI 1 month ago(2026-07-09)
GPT-5.6 TerraOpenAI 1 month ago(2026-07-09)
Grok 4.5xAI 2 months ago(2026-07-08)
Kimi K2.7 CodeMoonshot AI 2 months ago(2026-06-12)
Claude Fable 5Anthropic 2 months ago(2026-06-09)
Claude Opus 4.8Anthropic 3 months ago(2026-05-27)
Qwen3.7 MaxAlibaba 3 months ago(2026-05-21)
Gemini 3.5 FlashGoogle 3 months ago(2026-05-19)
Gemini 3.1 Flash LiteGoogle 4 months ago(2026-05-07)
GPT-5.5OpenAI 4 months ago(2026-04-24)
Grok 4.20xAI 5 months ago(2026-03-31)
MiniMax M2.7MiniMax 5 months ago(2026-03-18)
GPT-5.4 MiniOpenAI 5 months ago(2026-03-17)
GPT-5.4OpenAI 6 months ago(2026-03-05)
Gemini 3.1 Pro PreviewGoogle 6 months ago(2026-02-19)
Claude Sonnet 4.6Anthropic 6 months ago(2026-02-17)
Qwen3.5 Plus 2026-02-15Alibaba 6 months ago(2026-02-16)
GLM 5Z.ai 6 months ago(2026-02-11)
Claude Opus 4.6Anthropic 7 months ago(2026-02-04)
Kimi K2.5Moonshot AI 7 months ago(2026-01-27)
Gemini 3 Flash PreviewGoogle 8 months ago(2025-12-17)
DeepSeek V3.2DeepSeek 9 months ago(2025-12-01)
Claude Opus 4.5Anthropic 9 months ago(2025-11-24)
Claude Haiku 4.5Anthropic 10 months ago(2025-10-15)
Claude Sonnet 4.5Anthropic 11 months ago(2025-09-29)
DeepSeek V3.1 TerminusDeepSeek 11 months ago(2025-09-22)
Mistral Medium 3.1Mistral 1.0 years ago(2025-08-13)
GPT-5 MiniOpenAI 1.0 years ago(2025-08-07)

Catalogue dates are model-availability dates populated from OpenRouter. When that field is missing, a reviewed Artificial Analysis model reference page may supply the date. Unknown dates stay unknown rather than being inferred.

Release cadence, line by line

When each model line last shipped and how often new versions land — Opus vs. Sonnet, Gemini Pro vs. Flash, GPT mainline vs. mini. Hover a line's history for every release we track.

Model line Latest release How long ago Typical gap Releases tracked
GPT (mainline)OpenAI GPT-5.6 Terra 1 month ago
2026-07-09
~76 days 5
GrokxAI Grok 4.5 2 months ago
2026-07-08
~99 days 2
KimiMoonshotAI Kimi K2.7 Code 2 months ago
2026-06-12
~136 days 2
Claude OpusAnthropic Claude Opus 4.8 3 months ago
2026-05-27
~112 days 3
QwenQwen Qwen3.7 Max 3 months ago
2026-05-21
~94 days 2
Gemini FlashGoogle Gemini 3.5 Flash 3 months ago
2026-05-19
~141 days 3
GPT mini / nanoOpenAI GPT-5.4 Mini 5 months ago
2026-03-17
~222 days 2
Gemini ProGoogle Gemini 3.1 Pro Preview 6 months ago
2026-02-19
1
Claude SonnetAnthropic Claude Sonnet 4.6 6 months ago
2026-02-17
~141 days 2
GLMZ.ai GLM 5 6 months ago
2026-02-11
1
DeepSeekDeepSeek DeepSeek V3.2 9 months ago
2025-12-01
~70 days 2
Claude HaikuAnthropic Claude Haiku 4.5 10 months ago
2025-10-15
1

Every version bump counts as a release. Cadence is computed only from the current evaluation roster, so lines may skip releases outside it; “typical gap” is the median days between consecutive dated roster releases. We do not turn that limited history into a release forecast.

Release habits by lab

Within the dated models in our current roster: how often releases landed, and the most common release weekday in this limited sample.

OpenAI

  • Releases tracked: 7
  • Typical gap: ~76 days
  • Favourite day: Thursday (5 of 7)
  • Latest: GPT-5.6 Terra 1 month ago

Anthropic

  • Releases tracked: 7
  • Typical gap: ~40 days
  • Favourite day: Wednesday (3 of 7)
  • Latest: Claude Fable 5 2 months ago

Google

  • Releases tracked: 4
  • Typical gap: ~64 days
  • Favourite day: Thursday (2 of 4)
  • Latest: Gemini 3.5 Flash 3 months ago

Frequently asked

When did OpenAI last release a GPT (mainline) model in this roster?

GPT-5.6 Terra is the most recent GPT (mainline) release in Spring Prompt's current evaluation roster, released 1 month ago (2026-07-09).

How often has OpenAI released GPT (mainline) models in this roster?

Across the 5 dated GPT (mainline) models in the current evaluation roster, the median gap is about 76 days. This is a description of the roster, not a release forecast.

When did xAI last release a Grok model in this roster?

Grok 4.5 is the most recent Grok release in Spring Prompt's current evaluation roster, released 2 months ago (2026-07-08).

When did MoonshotAI last release a Kimi model in this roster?

Kimi K2.7 Code is the most recent Kimi release in Spring Prompt's current evaluation roster, released 2 months ago (2026-06-12).

When did Anthropic last release a Claude Opus model in this roster?

Claude Opus 4.8 is the most recent Claude Opus release in Spring Prompt's current evaluation roster, released 3 months ago (2026-05-27).

How often has Anthropic released Claude Opus models in this roster?

Across the 3 dated Claude Opus models in the current evaluation roster, the median gap is about 112 days. This is a description of the roster, not a release forecast.

When did Qwen last release a Qwen model in this roster?

Qwen3.7 Max is the most recent Qwen release in Spring Prompt's current evaluation roster, released 3 months ago (2026-05-21).

When did Google last release a Gemini Flash model in this roster?

Gemini 3.5 Flash is the most recent Gemini Flash release in Spring Prompt's current evaluation roster, released 3 months ago (2026-05-19).

How often has Google released Gemini Flash models in this roster?

Across the 3 dated Gemini Flash models in the current evaluation roster, the median gap is about 141 days. This is a description of the roster, not a release forecast.

Continue exploring

Open a model’s evidence profile

Release recency is context, not capability. Use the model profiles to inspect source-specific benchmark records, configurations, and operational facts without turning age into a ranking.

Browse model profiles