Last checked November 19, 2026 · News roundup
Today in brief: OpenAI shipped GPT-6.2 Astra and Sol on November 17 and cut Astra's API price by 20%, a day before Google answered with Gemini 4 Neon at $3/$15 per million tokens. Anthropic priced its IPO at $83 a share, above the range, and ANT is set to start trading on Nasdaq today. On the business side: a $280 million round for Payhawk, Perplexity's $1 billion run rate, and a PlayStation Network meltdown hours before GTA VI unlocked.
GPT-6.2 price and access: Astra drops to $8/$40, Sol stays at $2/$10
OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17, 2026. Astra is the heavy model for software work and research; Sol is the cheaper everyday and agent model. In the API they are gpt-6.2-astra and gpt-6.2-sol.
What it costs: Astra lists at $8 per million input tokens and $40 per million output tokens, down from $10/$50 for GPT-6 Astra. Sol keeps GPT-6.1 Sol's $2/$10. According to Data Studios' breakdown of OpenAI's pricing pages, cached input on Astra falls to $0.40, prompts over 272K input tokens are billed at higher rates, and the Ultrafast tier costs 3x standard until February 28, 2027, then 6x from March 1.
Who gets it in ChatGPT: Plus, Pro, Enterprise and Edu users get both models in ChatGPT Work and Codex. An OpenAI staff member said on the developer forum that the rollout does not add either model to regular Chat. Pro 500 subscribers can run both in Ultrafast; Pro 200 gets Ultrafast starting with Sol. Business, Free and Go plans were not mentioned.
What to do: if you pay for Astra-class API calls, re-run your cost math now; the cut alone is 20% before caching. OpenAI's own system card also rates Astra as the first externally released model to hit its "High" threshold for AI self-improvement, and both models face tighter limits on cyber requests. For how Pro usage caps worked before this release, see our breakdown of ChatGPT Pro usage limits.
| Model | Input | Output | Note |
|---|---|---|---|
| GPT-6.2 Astra | $8.00 | $40.00 | Was $10 / $50 on GPT-6 Astra |
| GPT-6.2 Sol | $2.00 | $10.00 | Unchanged from GPT-6.1 Sol |
| Gemini 4 Neon | $3.00 | $15.00 | New Nov 18; cached input $0.15 |
| Gemini 4 Argon / Carbon | $2.00 → $4.00 | $10.00 → $20.00 | Promo ends Nov 30; new rate from Dec 1 |
| Claude Opus 5.5 | $4.00 | $20.00 | As cited by VentureBeat |
Gemini 4 Neon: Carbon-level quality for AI Pro subscribers, $3/$15 in the API
Google launched Gemini 4 Neon on November 18, the third Gemini 4 model this month after Carbon (Nov 10) and Hydrogen (Nov 12). Google says Neon matches Carbon on coding, reasoning and agent tasks at Flash-like speed.
Who gets it: it is rolling out in the Gemini app for Google AI Pro and AI Ultra subscribers, and in the paid Gemini API, Google AI Studio, Android Studio and Antigravity. Per Google's Logan Kilpatrick, as quoted by VentureBeat, Pro users can pick Neon while Carbon stays Ultra-only. 9to5Google reports Neon replaces Argon in the app's model picker and Gemini 3.1 Pro is being retired; the new picker reads Auto, Everyday help (Hydrogen), Deeper analysis (Neon), and Code & complex work (Carbon).
Benchmarks, per VentureBeat: 61 on the Artificial Analysis Intelligence Index versus 59 for Claude Opus 5.5, and 46.2% on FrontierCode 2.0 versus Opus 5.5's 47.3%. Treat these as third-party snapshots.
The price angle: Neon costs 25% less per token than Argon and Carbon will from December 1, when their $2/$10 launch promo ends and rates double to $4/$20. If you built on Argon, our Gemini 4 Argon price and access guide has the full background; the update is that Neon is now the cheaper path for most of that work.
MAI Code 2 in GitHub Copilot: Microsoft's own coding model, at a promo price
At Ignite on November 17, Microsoft AI introduced MAI Code 2 and MAI Code 2 Flash, in-house coding models aimed at the same jobs as OpenAI's Sol and Anthropic's Opus. Both are available in GitHub Copilot now. Code 2 Flash is also in Microsoft Foundry as a limited-access preview, and Microsoft says OpenRouter, Vercel, OpenCode and Kilo are "coming soon."
Microsoft did not publish dollar prices. Its claims are relative: averaged across four coding benchmarks, MAI Code 2 reaches 98% of GPT-6.1 Sol's score at 40% of its cost, and 96% of Claude Opus 5.5's at 5% of its cost. Both cost figures include a 50% promotional discount, so expect them to roughly double once the promo ends. In blind front-end tests with 258 annotators, developers preferred MAI Code 2's output over GPT-6 Sol 64% of the time, Microsoft says.
Who it affects: Copilot users who pick models manually, and teams paying per token in Foundry. Microsoft also announced MAI Work Flash for Microsoft 365 tasks. What to do: try MAI Code 2 on routine refactors in Copilot this week while the promo pricing applies, and keep your current model for anything security-sensitive until independent results arrive.
ChatGPT for Sales: a Sales plugin that reads Salesforce and HubSpot
OpenAI added a dedicated Sales plugin to ChatGPT this week, alongside more than 30 new or expanded partner plugins, according to its announcement. The plugin page lists skills such as account plans, meeting briefs, competitive briefs, call follow-ups, contact enrichment and pipeline reviews that flag at-risk deals.
Named connections include Salesforce, HubSpot, Slack, Microsoft Teams, Zoom, Granola, Fireflies and Otter.ai, plus 19 more. The plugin can both read and write, which means it can update CRM records, not only summarize them.
Cost and access: OpenAI does not list a separate price. It says availability "depends on the plugin, your plan, and workspace settings," and some connections need an admin to approve them. Expect Business and Enterprise workspaces to see it first.
What to do: if you run a small sales team, ask your admin to switch on the CRM connector in a test workspace before letting it write to live deals. For prompts that already work without the plugin, our ChatGPT prompts for sales reps still apply, and the sales outreach tools hub covers the prospecting tools that feed a CRM in the first place.
Anthropic IPO price: $83 a share, with ANT set to trade on Nasdaq today
Update to our earlier IPO coverage: Anthropic priced its offering late on November 18 at $83 per share, above the $70–$78 range it set the week before. It sold 980 million Class A shares: about 956 million new shares from the company and about 24 million from existing holders.
That is roughly $81 billion in gross proceeds before fees, of which about $79 billion goes to Anthropic. Underwriters Goldman Sachs and Morgan Stanley have a 30-day option on 147 million more shares; if fully used, the deal reaches about $93.5 billion. CNBC called it the largest offering on record, ahead of SpaceX.
When it trades: shares are expected to start trading on the Nasdaq Global Select Market under the ticker ANT on Thursday, November 19, with no fixed time for the first trade. The deal is expected to close on Friday, November 20.
What it means for Claude users: nothing changes in plans or prices today. A listed Anthropic will report revenue and usage each quarter, which makes Claude pricing moves more visible. This is news, not investment advice; first-day IPO prices can swing sharply.
Is PSN down? What happened to PlayStation Network before GTA VI
PlayStation Network sign-in problems spiked on the evening of November 18, hours before Grand Theft Auto VI unlocked at midnight local time. WRAL reported more than 1,500 Downdetector reports, while PlayStation's own status page still showed all services up. Xbox Live users also reported trouble, according to Notebookcheck.
Sony gave no cause. Polygon, as cited by Notebookcheck, pointed to server overload from millions of consoles downloading the roughly 212 GB game at the same time. Players also need a PS5 system update of about 1.3 GB, released Tuesday, before the game will run.
What to do if it fails for you: install the system update first, leave the console in rest mode to finish the download, and check the official PlayStation status page before resetting anything. Launch-rush load problems tend to ease as the queue clears; a fault that persists after the status page turns yellow or red is Sony's to fix, not your router's.
Why it is in a SaaS roundup: like the Apple and Chase outages earlier this week, it is a reminder that "is it down?" checks should start with the vendor's status page and Downdetector, not support tickets.
Payhawk funding: $280 million at a $2 billion valuation for AI spend agents
Payhawk, a spend management platform that competes with Pleo, Moss and Spendesk, announced a $280 million Series C on November 18 at a $2 billion valuation, TechCrunch reports. Goldman Sachs' growth equity arm led the round, with Greenoaks, Lightspeed and QED joining. Total funding now tops $500 million.
Founded in 2018 and operating in 33 countries, Payhawk sells a finance platform that uses AI agents to help companies manage expenses and procurement. The company says it still had 70% of the $100 million it raised in 2022, which suggests the new money is for expansion rather than survival. It did not disclose revenue, customer counts or US-specific plans.
Who should care: US finance teams comparing Ramp or Brex with a European-born rival. A well-funded challenger usually means sharper pricing in sales negotiations, so it is a good moment to ask your current vendor about multi-entity and AI-agent features before renewal.
Perplexity revenue: a $1 billion run rate, ahead of its own target
Perplexity CEO Aravind Srinivas said on November 16 that the company has crossed a $1 billion annualized revenue run rate. According to Runtime Wire, revenue was under $250 million at the start of 2026 and passed $750 million by August, so the company beat the $750 million year-end goal it had set.
Srinivas added on X that Perplexity has "strong margins and a healthy growth rate," without publishing figures. Run rate means current monthly revenue multiplied by 12, not booked annual revenue.
Why it matters for subscribers: Perplexity did not split the number by plan, but fast revenue growth usually means more features land in paid tiers first. If you are weighing Pro or Max, judge them on today's features rather than waiting for a price cut.
Saazora take: three model launches in two days, and the price story beats the benchmark story. Astra got cheaper, Neon undercuts Gemini's own December prices, and Microsoft is pricing to win Copilot users. If your team pays per token, this is the week to re-quote your workloads.
Reader questions on today's news
Can I use GPT-6.2 in regular ChatGPT chat?
Not yet. OpenAI is rolling GPT-6.2 Astra and Sol out in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu subscribers. An OpenAI staff member said on the developer forum that this release does not add either model to regular Chat. Business, Free and Go plans were not mentioned in the announcement.
Is Gemini 4 Neon free?
No free tier was announced. Neon is rolling out in the Gemini app for Google AI Pro and AI Ultra subscribers, and in the paid Gemini API at $3 per million input tokens and $15 per million output tokens. Google AI Studio and Android Studio also list it. None of the launch coverage mentions access on the free Gemini plan.
What time does Anthropic stock start trading?
Anthropic did not set a time. Shares are expected to begin trading on the Nasdaq Global Select Market under the ticker ANT on Thursday, November 19, 2026. Large IPOs often open after midday Eastern time, once underwriters balance buy and sell orders, so the first trade could come well after the 9:30 a.m. ET open.
Does ChatGPT for Sales cost extra?
OpenAI has not listed a separate price for the Sales plugin. It says availability depends on your ChatGPT plan, the plugin and workspace settings, and some CRM connections need admin approval. In practice, that points to Business and Enterprise workspaces first. Check your workspace's plugin settings before buying a new seat for it.
Sources: OpenAI Developer Community, Nov 17, 2026 · Data Studios, Nov 2026 · 9to5Google, Nov 18, 2026 · VentureBeat, Nov 18, 2026 · Microsoft AI, Nov 17, 2026 · OpenAI Sales plugin page, checked Nov 19, 2026 · investingLive, Nov 18, 2026 · StartupHub, Nov 18, 2026 · CNBC, Nov 18, 2026 · WRAL, Nov 18, 2026 · Notebookcheck, Nov 18, 2026 · TechCrunch, Nov 18, 2026 · Runtime Wire, Nov 17, 2026