33 events in a year. Then 155 in the next.
Every model-market event Vitna has recorded since Mar 2023, newest first. Each entry links to a primary source on a domain the provider itself publishes on, and carries the date that source was last fetched; the newest of those checks is 2026-11-19. The log only ever gains entries.
Every event at its true date, newest left.
Each move as a share of the price before it.
83 detected moves carry both of their periods, so each bar is the move as a share of the price before it: cuts run left, rises run right, and the longest bar is the largest move. Each model leads to its own page and each meter to the move's own page. Curated entries cite an announcement rather than two readings and list below without a bar.
Every entry, with the document behind it.
42 of 216 events. OpenAI.
- 2026OpenAIpricesource ↗
GPT-5.6 Sol cut 20 percent on input and 33 percent on output
The frontier tier follows Luna and Terra down, six weeks after the generation launched.
- OpenAIavailabilitysource ↗
Daybreak Blue and Daybreak Red open to approved defenders
Two access tiers released to vetted customers rather than to the API generally.
- OpenAIpricesource ↗
GPT-5.6 Terra list price cut 20 percent
The balanced tier is cut on the same day as Luna, and Fast mode replaces Priority processing.
- OpenAIpricesource ↗
GPT-5.6 Luna list price cut about 80 percent
Three weeks after launch, Luna drops from $1.00 / $6.00 to $0.20 / $1.20 per million tokens.
- OpenAIdeprecationsource ↗
Nine legacy audio and realtime models are deprecated
Scheduled for 20 January 2027, with GPT-Realtime-2.1 and GPT-Audio-1.5 named as the replacement.
- OpenAIlaunchsource ↗
GPT-5.6 Sol, Terra and Luna launch
Luna, the fast tier, lists at $1.00 / $6.00 per million tokens.
- OpenAIdeprecationsource ↗
The GPT-5 and o3 snapshots are deprecated
Six models scheduled for 11 December 2026, sixteen months after the GPT-5 family launched.
- OpenAIlaunchsource ↗
GPT-5.5 and GPT-5.5 Pro reach a million tokens
Both support a 1M token context window.
- OpenAIdeprecationsource ↗
Thirteen legacy GPT snapshots are deprecated at once
Scheduled for 23 October 2026, with the three GPT-5.6 tiers named as the replacement.
- OpenAIlaunchsource ↗
GPT-5.4 mini and nano complete the ladder
The small and smallest tiers land twelve days after the frontier pair.
- OpenAIlaunchsource ↗
GPT-5.4 and GPT-5.4 Pro released
Tool search, computer use and compaction arrive with them.
- 2025OpenAIlaunchsource ↗
GPT-5.2 released with an extra reasoning step
Adds an xhigh reasoning effort and client-side compaction.
- OpenAIlaunchsource ↗
GPT-5.1 released with a codex pair beside it
Improved steerability and agentic work, with gpt-5.1-codex and gpt-5.1-codex-mini.
- OpenAIlaunchsource ↗
GPT-5 Pro tops the GPT-5 ladder
Released with a mini realtime tier, gpt-audio-mini, gpt-image-1-mini and the Sora 2 video API.
- OpenAIdeprecationsource ↗
GPT-3.5 Turbo Instruct and the base models are deprecated
Scheduled for 28 September 2026, with GPT-5.6 Terra named as the replacement for davinci-002, babbage-002 and the last GPT-3.5 snapshots.
- OpenAIdeprecationsource ↗
The Assistants API is deprecated
Scheduled for 26 August 2026, with the Responses and Conversations APIs named as the replacement. A year of notice for an interface rather than a model.
- OpenAIlaunchsource ↗
The GPT-5 family launches with three tiers
gpt-5, gpt-5-mini and gpt-5-nano ship together, with a minimal reasoning effort setting.
- OpenAIavailabilitysource ↗
GPT-4.5 preview is shut down
Four and a half months of life from research preview to retirement, the shortest the record carries.
- OpenAIpricesource ↗
o3 list price cut across every request type
The reduction lands on the same day o3-pro is released above it.
- OpenAIdeprecationsource ↗
o1-preview and o1-mini are deprecated
The models that opened the reasoning tier are scheduled for 28 July and 27 October 2025, nineteen months after o1-preview shipped.
- OpenAIlaunchsource ↗
o3 and o4-mini join the reasoning line
Two reasoning models released with the first Codex CLI.
- OpenAIdeprecationsource ↗
GPT-4.5 preview is deprecated on the day GPT-4.1 launches
The research preview is scheduled for removal in the same changelog entry that announces its replacement.
- OpenAIlaunchsource ↗
GPT-4.1 launches as a three-model ladder
gpt-4.1, gpt-4.1-mini and gpt-4.1-nano land on the same day.
- OpenAIlaunchsource ↗
GPT-4.5 enters research preview
A preview tier that is deprecated seven weeks later, the shortest life the record carries.
- OpenAIlaunchsource ↗
o3-mini opens a cheaper reasoning tier
The reasoning line gains a small model five months after o1-mini.
- 2024OpenAIlaunchsource ↗
o1 reaches general availability with realtime and audio models
The reasoning tier leaves preview alongside new realtime and audio snapshots and WebRTC support.
- OpenAIpolicysource ↗
Prompt caching, distillation and the Realtime API arrive together
Caching lands on the same day as the Realtime API, model distillation, image fine-tuning and Evals.
- OpenAIavailabilitysource ↗
The 2023 GPT-3.5 Turbo snapshots are shut down
gpt-3.5-turbo-0301, 0613 and the 16k variant stop answering on the same day, on the schedule set in June and November 2023.
- OpenAIlaunchsource ↗
o1-preview and o1-mini open the reasoning tier
A model class billed for the tokens it thinks with, priced apart from the GPT line.
- OpenAIlaunchsource ↗
GPT-4o mini opens the cost-efficient tier
A small omni model priced for volume; fine-tuning follows five days later.
- OpenAIavailabilitysource ↗
GPT-4 0314 is shut down
The first pinned GPT-4 snapshot stops answering, one year to the day after its notice.
- OpenAIdeprecationsource ↗
GPT-4 32k and the vision preview are deprecated
gpt-4-vision-preview is scheduled for 6 December 2024 and gpt-4-32k for a year later, after vision folded into the main model.
- OpenAIlaunchsource ↗
GPT-4o released in the API
The omni tier lands as a single model for text, vision and audio.
- OpenAIpolicysource ↗
The Batch API arrives and halves the rate for patient work
Asynchronous request processing at a discount to the synchronous rate, the first time latency is priced separately from capability.
- OpenAIlaunchsource ↗
GPT-4 Turbo with Vision reaches general availability
The preview tier becomes a generally available model with image input in the same request.
- OpenAIlaunchsource ↗
A dated GPT-3.5 Turbo snapshot joins the API
gpt-3.5-turbo-0125 is pinned as its own model id, and is still listed on the pricing index today.
- OpenAIavailabilitysource ↗
The pre-chat generation is shut down
InstructGPT, the base models, the edits endpoint and the first-generation embeddings stop answering, six months after the notice.
- 2023OpenAIlaunchsource ↗
GPT-4 Turbo opens the long-context tier
The first DevDay release: GPT-4 Turbo in preview alongside an updated GPT-3.5 Turbo, vision input, the Assistants API, DALL-E 3 and text to speech.
- OpenAIdeprecationsource ↗
GPT-3.5 Turbo 0613 is deprecated on the day GPT-4 Turbo launches
gpt-3.5-turbo-0613 and its 16k sibling are scheduled for 13 September 2024, announced in the same DevDay that opens the tier above them.
- OpenAIdeprecationsource ↗
InstructGPT, the base models and the first embeddings are deprecated
text-davinci-003, the bare ada and davinci models, the edits endpoint and every first-generation embedding are scheduled for 4 January 2024. The whole pre-chat generation, in one notice.
- OpenAIdeprecationsource ↗
The first pinned GPT-4 and GPT-3.5 snapshots are deprecated
gpt-4-0314 is scheduled for 13 June 2024 and gpt-3.5-turbo-0301 for 13 September 2024, the first time a dated snapshot is given a year to live.
- OpenAIdeprecationsource ↗
The Codex models are deprecated with three days of notice
code-davinci-002, code-cushman-002 and their predecessors are scheduled for shutdown on 23 March. The shortest notice period in the record, by a factor of twenty.
An event appears here only with a primary-source URL verified on the date the entry carries. Secondary coverage can corroborate a date but never stands as the source. A detected price move is published only after a second read at least an hour later agrees, and cites the archived document either side of it.
A week with 3 events or more is a page of its own, with its counts, its sized moves and every entry in it. The newest is the week of 16 Nov 2026, with 16 price moves, and all 16 are listed by week.
This log is published as a JSON Feed at /api/v1/feed.json, public and with no key, carrying every entry here and the decisions the engine has proposed against them, and as RSS at /log/rss.xml for a feed reader, with the same items under the same ids. Every headline above is also a page of its own, at an address that keeps pointing at it. Nothing in either feed is a recommendation to act.
