In this guide · 8 sections
- 01What launched in September and October 2026
- 02How the new models rank on published leaderboards
- 03Price changes worth knowing
- 04Models being retired: check your chatbot and automations
- 05Should an SME switch? Our verdict by situation
- 06What we are watching next quarter
- 07Do it yourself or get help?
- 08Frequently asked questions
Short answer
Most SMEs using ChatGPT, Claude or Gemini through a subscription get the new models automatically and need to do nothing. If you have a website chatbot or automation built on an API, check the model name: Claude Sonnet 4.5 retires on 30 November 2026 and Gemini 3.7 Flash is already redirected.
What launched in September and October 2026
Dates and prices come from each vendor's own announcements and release notes. API prices are per million tokens (a token is a short chunk of text), in US dollars, as of October 2026.
| Date | Vendor | Model | API price (in / out) |
|---|---|---|---|
| 3 Sep | OpenAI | GPT-6 Astra, flagship | US$10 / US$50 |
| 22 Sep | OpenAI | GPT-6 Sol and GPT-6 Luna, cheaper tiers | Luna US$0.10 / US$0.50 |
| 22 Sep | Anthropic | Claude Opus 5.5 | US$4 / US$20 |
| 28 Sep | Anthropic | Claude Sonnet 5.5 | US$2 / US$10 |
| 29 Sep | OpenAI | GPT-6.1 Sol, near-Astra results at lower cost | US$2 / US$10 |
| 7 Oct | Anthropic | Claude Haiku 5.5, fast and cheap | US$0.10 / US$0.50 (prompts up to 100K tokens) |
| 13 Oct | Gemini 4 Argon, generally available | US$2 / US$10 until 30 Nov, then US$4 / US$20 |
Anthropic also released Claude Fable 5.1 on 1 September at US$10 / US$50, aimed at long-running agent work. Google first announced Gemini 4 Argon on 30 September for a group of trusted cyber defenders before the general release.
How the new models rank on published leaderboards
IT Sorted did not run these tests. These are third-party leaderboards, with the date we read them.
- Artificial Analysis Intelligence Index (read 18 October 2026): Claude Opus 5.5 at maximum effort scores 58, the highest listed, followed by Claude Sonnet 5.5 at 56. GPT-6 Astra and Gemini 4 Argon both score 53, and GPT-6.1 Sol 52. See the Artificial Analysis leaderboard.
- Arena text leaderboard (dated 15 October 2026), which ranks by human votes: Claude Opus 5.5 and Gemini 4 Argon share first place for instruction following, and Gemini 4 Argon leads for longer prompts. Argon's scores are marked preliminary because it is new. See Arena.
- Cheaper models are close behind. GPT-6.1 Sol and Claude Sonnet 5.5 cost a fifth of GPT-6 Astra's per-token price and score at or near the top models on these indexes.
For a side-by-side of the business plans, seat prices and data terms, see ChatGPT vs Claude vs Gemini for business.
Price changes worth knowing
Several vendors cut prices this quarter, which lowers the cost of running a website chatbot or an email-sorting automation.
- Claude Opus 5.5 is priced at US$4 / US$20, below Claude Opus 5 at US$5 / US$25 (Anthropic release notes).
- Claude Sonnet 5 kept its introductory US$2 / US$10 price permanently; a planned rise on 1 September was cancelled.
- Claude Sonnet 5.5 cached-input reads dropped from US$0.20 to US$0.10 per million tokens on 7 October.
- OpenAI says GPT-6 Sol and Luna API prices are 50% lower than GPT-5.6 promotional pricing.
- Gemini 4 Argon is at an introductory price until 30 November 2026. It doubles to US$4 / US$20 from 1 December, so budget for the higher rate.
- ChatGPT Business added Premium seats in August 2026 at US$100 per user per month billed annually, alongside the US$20 standard seat.
Models being retired: check your chatbot and automations
If someone built a chatbot, quote tool or automation for you that calls an AI model directly, it names a specific model. When that model is retired, the feature can stop working.
| Model | What happens | Date |
|---|---|---|
| Claude Sonnet 4.5 | Retired on the Claude API; Anthropic recommends Sonnet 5.5 | 30 November 2026 |
| Claude Opus 4.1 | Already retired; requests return an error | 5 August 2026 |
| Gemini 3.7 Flash | Deprecated; requests are routed to Gemini 3.8 Flash | 8 October 2026 |
| Gemini 3.5 Flash | Deprecated; requests are routed to Gemini 3.6 Flash | 8 October 2026 |
Should an SME switch? Our verdict by situation
| You use | Do you need to act? | What to do |
|---|---|---|
| ChatGPT, Claude or Gemini subscriptions only | No | New models appear in the model picker. Tell staff which one to use for what. |
| A website chatbot on Claude Sonnet 4.5 | Yes, before 30 Nov | Move to Sonnet 5.5 or Haiku 5.5 and retest your answers. |
| A chatbot or automation on Gemini 3.5 or 3.7 Flash | Soon | It is already redirected. Update the model name and check answer quality and cost. |
| High-volume email sorting or FAQ replies | Optional | Try a cheaper tier such as Claude Haiku 5.5 or GPT-6 Luna to cut cost. |
| Long documents, tenders or contracts | Optional | Trial Claude Opus 5.5 or Gemini 4 Argon on a sample document. |
If your business uses open-weight models such as DeepSeek or Qwen, the questions are different. Our guide on whether DeepSeek or Qwen is safe for business covers data location and PDPA.
What we are watching next quarter
A few items from this quarter may still change. OpenAI's GPT-6.1 Astra did not launch at DevDay; TechCrunch reported that the release was scrapped over safety concerns raised in internal testing. Google's Argon pricing changes on 1 December. Anthropic's Claude Max and Team plans now include monthly API credits, which may make small automations cheaper for Team customers.
We update this roundup each quarter using the same sources, so figures stay comparable. The next edition is due in January 2027.
Do it yourself or get help?
If you only use AI through subscriptions, there is nothing to do except tell staff which model to pick. If you have a chatbot or automation that calls an API, someone needs to find the model name in the code or settings, test the replacement and switch it before the retirement date.
AI Website Assistant (S$790) builds a website assistant on your own accounts with a model you choose, and Workflow Automation (S$390 per workflow) sets up email sorting or quote automations. Both are documented so you can see which model is used and change it later. Our guide to AI chatbots for your website covers costs and PDPA risks.
Frequently asked questions
Anthropic released Claude Haiku 5.5 on 7 October 2026, and Google made Gemini 4 Argon generally available on 13 October. Late September also brought Claude Opus 5.5 and Sonnet 5.5 from Anthropic and GPT-6.1 Sol from OpenAI. GPT-6 Astra launched earlier, on 3 September.
Usually not. Subscription plans such as ChatGPT Business, Claude Team and Gemini in Workspace add new models automatically. You only need to act if a website chatbot or automation calls a specific model through an API, because retired models stop working or get redirected.
Anthropic announced on 30 September 2026 that Claude Sonnet 4.5 will be retired on the Claude API on 30 November 2026, and recommends migrating to Claude Sonnet 5.5. Test the new model before switching, because some request settings that worked on older models now return errors.
Google lists an introductory API price of US$2 per million input tokens and US$10 per million output tokens until 30 November 2026, rising to US$4 and US$20 from 1 December 2026. In the Gemini app it is available to Google AI Pro and Ultra subscribers. It is not on the free API tier.
On the Artificial Analysis Intelligence Index read on 18 October 2026, Claude Opus 5.5 scored highest. On Arena's human-vote leaderboard of 15 October, Claude Opus 5.5 and Gemini 4 Argon tied for first in instruction following. For most SME tasks the gap between the top models is small.