AI DevelopmentIndustry Guide5 min readPublished September 2, 2026

Dated on the vendor’s own announcement, not on a listing

AI Model Releases: September 2026 Tracker and Dated Ledger

Three frontier releases in the first two days of the month, a gated cyber model, a contributor tier, and six late-August releases that set the prices September is being judged against. Every row carries the vendor’s own date and primary; OpenRouter listing times are corroboration, never the date. Refreshed in place through September 30.

DA
Digital Applied Team
Senior strategists · Published Sep 2, 2026
PublishedSep 2, 2026
Read time5 min
As ofSep 3, 2026
September releases logged
6
Sep 1–2, 2026, each with a vendor primary
Late-August carry-ins
6
Aug 25–31; the prices this month is measured against
OpenRouter listings since Aug 26
14
live API pull, Sep 3; includes batch surfaces and aliases
Dated events still ahead
4
Sep 8 discount end, Sep 14 limit change, fall, Dec 31

September 2026 began with the densest 48 hours of frontier releases since the August wave. Anthropic shipped Claude Fable 5.1 and Mythos 5.1 on September 1 at an unchanged list price with three breaking API changes. On September 2 Google released Gemini 3.8 Flash at the same introductory price as 3.7 Flash, with the end of that price now printed, alongside a Fairwind-gated Cyber variant; and Meta released Muse Spark 1.3 with a contributor tier listed the same evening.

This ledger is the September edition of the dated tracker we have kept since May and August. Each row is dated on the vendor’s own announcement, priced from the vendor or the OpenRouter listing where the vendor publishes no price, and linked to the post where we covered it. Releases after September 3 are added to this page rather than given new posts, unless one forces a decision on its own.

Key takeaways
  1. 01
    Three frontier models in two days, none of them a price cut.Fable 5.1 keeps Fable 5’s $10 and $50; only cache reads fell. Gemini 3.8 Flash keeps 3.7’s $0.75 and $3.75 and now carries a January doubling. Muse Spark 1.3 keeps 1.2’s $1.25 and $4.25. The month’s cheaper rows are all late-August carry-ins.
  2. 02
    Two of the six September rows are gated or conditional.Mythos 5.1 is available only through Anthropic’s verification programmes, and Gemini 3.8 Flash Cyber only through Google’s new Fairwind Program. Neither has a public price.
  3. 03
    Listing dates and launch dates disagree by up to six days this month.IBM’s Granite 4.2 8B carries an August 25 release date on its model card and an August 31 OpenRouter listing. The frontier releases listed within hours. The rule stands: the vendor’s date is the date.
  4. 04
    Four dated events are already fixed for the rest of the month and beyond.Mercury 2.5’s 80% discount ends September 8 at 07:00 UTC; Claude Code’s weekly-limit change lands September 14; Anthropic’s Enterprise Frontier Safeguards are “this fall”; Gemini 3.8 Flash’s intro price expires December 31.

01The ledgerThe September ledger.

AI model releases dated September 1–2, 2026. Dates are the vendor’s announcement dates; OpenRouter listing times are recorded as corroboration. Vendor-run benchmark figures are labelled. Prices per million tokens, input / output.
DateVendor · modelWhat it isPriceContextPrimary sourceReplaces
September 1–2 · verified against vendor primaries
Sep 1Anthropic · Claude Fable 5.1 and Claude Mythos 5.1Same model, two safeguard levels; Mythos 5.1 gated to Anthropic’s trusted-access programmes. Three breaking API changes: forced tool choice removed, thinking bound one way by model, history edits invalidate thinking (enforced for accounts created on or after Aug 31)$10 / $50 per M, unchanged from Fable 5; cache reads cut to $0.25 (was $1.00); batch $5 / $251M in, 128K outanthropic.com announcement; platform.claude.com migration guide. OpenRouter listed Sep 1, 18:03 UTCFable 5 (still available)
Sep 1Perplexity · Hybrid Compute on Mac, with PPLX Qwen 3.8 27BComputer tasks start in the cloud; sensitive steps and private-file access run locally on Apple silicon (macOS 15+, 24 GB minimum, 32 GB recommended) behind an on-device PII classifier; local work uses no cloud creditsIncluded in Perplexity Pro, Max and EnterpriseLocal model; not applicableperplexity.ai/hub (returns 403 to fetchers); tier-one same-day coverage used, as stated in our postCloud-only Computer tasks
Sep 2Google · Gemini 3.8 FlashThird Flash release in six weeks by Google’s count. HLE-Verified 54.9% (vendor-run); Artificial Analysis index 59, cost per task $0.58 vs $0.40 for 3.7 Flash at the same per-token price. Live in the API, AI Studio, Antigravity, Gemini app, AI Mode and Sheets$0.75 / $3.75 per M to Dec 31, 2026; $1.50 / $7.50 from Jan 1, 2027 (Google’s footnote)1,048,576 (OpenRouter)blog.google launch post. OpenRouter listed Sep 2, 15:14 UTC3.7 Flash “remains fully supported”
Sep 2Google · Gemini 3.8 Flash CyberCyber variant with looser mitigations, gated through the new Fairwind Program for governments, critical-infrastructure operators and core technology platforms. CWE-Bench pass@1 47.2% and other figures are vendor-runNot publishedNot publishedblog.google launch post; deepmind.google/fairwind-program3.5 Flash Cyber (July)
Sep 2Meta · Muse Spark 1.3Asks clarifying questions, asks for help when stuck, confirms before consequential actions. Meta: ~20% fewer tool calls, ~25% fewer tokens than 1.2 (vendor-run). Artificial Analysis: index 61 at xhigh, $0.55 per task, ~57% more input tokens per task. Max reasoning “coming shortly”$1.25 / $4.25 per M, cache hit $0.15, unchanged1,048,576 (OpenRouter)research.meta.ai launch post; artificialanalysis.ai. OpenRouter listed Sep 2, 19:45 UTCMuse Spark 1.2 (August)
Sep 2Meta · Muse Spark 1.3 ContributorDiscounted tier on which Meta trains on your traffic; the same arrangement as the 1.2 contributor tier$0.10 / $0.20 per M, cached input $0.0021,048,576 (OpenRouter)OpenRouter listing Sep 2, 20:38 UTC (listing date; Meta’s launch post does not name the tier)Muse Spark 1.2 Contributor

Each September row has a post behind it. The Anthropic row’s breaking changes are in our Fable 5.1 cost and breaking-changes post; the Gemini row’s January price is the subject of its own post; the Muse row’s two conflicting token numbers are explained in the Muse Spark 1.3 post; and the Perplexity row is covered in our Hybrid Compute post, which states that Perplexity’s own hub page could not be fetched and that tier-one same-day coverage was used instead.

02ContextLate-August carry-ins.

Six releases in the last week of August are the price floor September is being read against, and two of them still have live dates inside September. They are recorded here once, dated on the vendor where a vendor date exists, so that a reader comparing Gemini 3.8 Flash’s $0.75 against GLM-5.3-Flash’s $0.075 has both rows on one page.

Releases dated August 25–31, 2026 that carry into September’s pricing picture. Dates are the vendor’s where published; where only an OpenRouter listing time exists, the row says so. Prices per million tokens, input / output.
DateVendor · modelWhat it isPriceContextPrimary sourceReplaces
August 25–31 · carry-ins
Aug 25IBM · Granite 4.2 8BDense 8B open model with reasoning effort levels; release date per IBM’s Hugging Face model card$0.10 / $0.15 per M on OpenRouter131,072huggingface.co/ibm-granite/granite-4.2-8b. OpenRouter listed Aug 31, 20:06 UTC, six days after the card’s release dateNot stated on the model card
Aug 26Z.ai · GLM-5.3-FlashMIT-licensed open-weight Flash model, revealed as the stealth ox-alpha model$0.075 / $0.25 per M on OpenRouter1,310,720z.ai/blog/glm-5.3-flash. OpenRouter listed Aug 26, 13:59 UTCGLM-5.2-class Flash
Aug 26Alibaba · Qwen3.8-FlashFlash-class release covered in our August 26–31 days$0.15 / $0.47 per M on OpenRouter1,000,000OpenRouter listed Aug 26, 19:37 UTC; vendor coverage in the August ledgerEarlier Qwen Flash tier
Aug 28Tencent · Hy4 previewPreview of Tencent’s next flagship; covered in the August ledger$0.834 / $2.501 per M on OpenRouter1,048,576OpenRouter listed Aug 28, 06:09 UTCHy3
Aug 28Z.ai · GLM-5.3 (753B) open weightsWeights released under a bespoke licence after a two-week hold Z.ai attributed to safety evaluation of faster-than-expected cyber capabilityOpen weights; hosted pricing variesPer hostz.ai/blog/glm-5.3GLM-5.2
Aug 31Inception · Mercury 2.5 PreviewDiffusion language model; 1,107 tokens per second claimed (vendor-run). Comparison set named by Inception: GPT-5.6 Luna (Low), Gemini 3.5 Flash-Lite, Claude Haiku 4.5$0.04 / $0.15 per M at an 80% discount until Sep 8, 07:00 UTC; list $0.20 / $0.75260,000 in, 65,536 outInception site; OpenRouter listed Aug 31, 20:57 UTCMercury 2 (February)

03CalendarDated events still to come.

Four events already have dates. Two are price changes, one is a usage-limit change covered in our Claude Code September 14 post, and one is a safeguards product Anthropic has announced but not shipped.

Sep 8, 07:00 UTC
Mercury 2.5 discount ends
$0.04 / $0.15 → $0.20 / $0.75

Inception’s 80% launch discount on Mercury 2.5 Preview expires. Anyone who routed a fast-cheap tier to it in the first week should have the list price in the forecast from this date.

Price change
Sep 14
Claude Code weekly limit change
Covered in our earlier post

The weekly-limit reduction Anthropic scheduled for September 14 lands. Details and the decision it forces are in our dedicated post; this row exists so the date is on the calendar.

Usage limits
Fall 2026
Anthropic Enterprise Frontier Safeguards
Announced, not shipped

Anthropic’s launch post says eligible customers can use Fable 5.1 under zero data retention until EFS arrives, phased “starting this fall.” Not available today; do not plan on it for a Q3 deployment.

Announced
Dec 31, 2026
Gemini 3.8 Flash intro price expires
$0.75 / $3.75 → $1.50 / $7.50 on Jan 1, 2027

Google’s launch-post footnote. The only frontier intro price this month with a published end date; the price index carries both numbers.

Price change
Announced without a date

Two items are real but undated and therefore not rows: Muse Spark 1.3’s max reasoning level, which Meta says is “coming shortly after we finish additional safety testing,” and the Muse Spark open-weights release, listed in Meta’s roadmap with no date and tracked on our withheld-weights ledger. They become rows when they get dates.

04AnalysisHow to read the month so far.

The pattern across the six September rows is that per-token prices held and the mechanisms around them moved. Anthropic changed the cache-read price and three API behaviours. Google left the price and published its expiry. Meta left the price and changed the model’s behaviour so that it reads more on a fixed task. A buyer comparing the month by list price sees nothing happening; a buyer comparing cost per task, which Artificial Analysis now publishes for every row above, sees Gemini 3.8 Flash up about 40% on its predecessor and Muse Spark 1.3 up from $0.40 to $0.55. The frontier model API price index carries the per-token rows; this ledger is where the dates live.

The second pattern is gating. Two of six September rows require an application, and both vendors’ gated models are the versions with cyber safeguards loosened. That is now the standing shape of a frontier launch, and the eligibility rules differ enough by lab that they need their own table rather than a column here. The third pattern is that the cheap end of the market is entirely open-weight or diffusion and entirely late-August: GLM-5.3-Flash, Qwen3.8-Flash, Granite 4.2 8B and Mercury 2.5 between them cover $0.04 to $0.15 per million input tokens. Whether September adds a frontier-lab entry at that price is the question the rest of this ledger will answer.

05MethodMethodology.

Methodology

A dated ledger, not a roundup: each row is a verified release with a vendor primary, and the page is refreshed in place rather than replaced by later instalments.

What was collected
Model releases with a vendor announcement dated September 1–30, 2026, plus late-August releases whose pricing or dated events carry into September. For each: date, vendor and model, what it is, price, context window, primary source, and what it replaces.
Sources
Vendor announcements (anthropic.com, platform.claude.com, blog.google, deepmind.google, research.meta.ai, z.ai, IBM’s Hugging Face model card, Inception); Artificial Analysis for independent index and cost-per-task figures; a live pull of the OpenRouter models API on September 3, 2026, filtered on created time, for listing times, prices and context lengths where the vendor publishes none.
As-of date
Collected September 3, 2026. The page is dated September 2 for the releases it opens with; this row is the only statement of the collection date, and it moves with each refresh.
Dating rule
The vendor’s announcement date is the row date. An OpenRouter listing time is corroboration only; it can lead or lag the real date by days or months, as the Granite row shows. Where no vendor date could be located, the row says so and gives the listing time.
Exclusions
Post-trains of open-weight models by third parties (they are listed on the host’s page, not here); models with no public listing or announcement; benchmark-only previews without a name.

06ConclusionSame prices, moved mechanisms.

September tracker

Two days, three frontier models, zero list-price cuts. The changes are in cache rates, token usage, expiry dates and who is allowed in.

The September ledger opens with the vendors holding their prices and changing everything around them: what a cache read costs, how many tokens a task uses, when an introductory price ends, and which customers get the version with the safeguards off.

This page will carry every dated September release as it lands, on the vendor’s date, with its primary. If a release forces a decision on its own it gets a post and a link from its row; if it does not, the row is the coverage.

Releases turned into decisions

Know what shipped, and what it changes.

We keep the model calendar, the price index and the routing rules in one place for the systems we run, so a September release becomes a forecast line and a router change rather than a surprise.

Free consultationExpert guidanceTailored solutions
What we work on

Model-portfolio engagements

  • Release monitoring with vendor-dated ledgers
  • Price-change calendars wired into forecasts
  • Router updates on each frontier release
  • Gated-model application support where eligible
  • Cost-per-task measurement on your own workload
FAQ · September 2026 releases

The questions we get about this month’s releases.

As of September 3: Claude Fable 5.1 and Claude Mythos 5.1 (Anthropic, September 1), Perplexity’s Hybrid Compute with the PPLX Qwen 3.8 27B local model (September 1), Gemini 3.8 Flash and the Fairwind-gated Gemini 3.8 Flash Cyber (Google, September 2), and Muse Spark 1.3 with a contributor tier (Meta, September 2). Each row above names the vendor primary.
Related dispatches

Continue exploring model releases and pricing.

AI Development

Grok 4.6 Lands: Same Price as 4.5, Frontier Parity

xAI's Grok 4.6 keeps Grok 4.5's headline rates, but the $2/$6 tier only holds below a 200K-token prompt: cross it and the whole request reprices.

August 12, 2026 · 13 minRead
AI Development

12 AI Models Released in One Week: March 2026 Guide

Twelve AI models launched in a single week of March 2026 from OpenAI, Google, Mistral, xAI, and more. Developer guide to capabilities, pricing, and selection.

March 15, 2026 · 13 minRead
AI Development

Who Gets the Cyber AI Models: Every Vetting Programme Listed

Google’s Fairwind joins Anthropic’s Glasswing and CVP, OpenAI’s Daybreak and Microsoft’s MDASH. One table of who is eligible for each cyber-capable model.

September 2, 2026 · 6 minRead
AI Development

Why an AI’s Reasoning Can’t Follow You to Another Model

Anthropic, OpenAI and Google now bind a model’s reasoning to the model that produced it. What each locks, what breaks on a switch, and how a router copes.

September 2, 2026 · 7 minRead
AI Development

AI Agent Memory 2026: Vector, Graph, Episodic Update

AI agent memory architectures compared after Code with Claude London — Anthropic Dreaming, Memory Tool, Google Memory Bank, vector, graph, episodic patterns.

May 24, 2026 · 16 minRead
AI Development

AI Agent Governance: Policy and Compliance 2026 Guide

AI agent governance framework for enterprises — access control, audit trails, data residency, and compliance with EU AI Act and SOC 2 requirements.

May 23, 2026 · 20 minRead