Quick Take
- Anthropic releases Claude Opus 4.8 globally on May 29, 2026, at the same price as Opus 4.7.
- The model is four times less likely to miss code errors than its predecessor, per Anthropic testing.
- Claude Mythos, now limited to select firms, will reach all users within weeks under Project Glasswing.
Claude Opus 4.8 launches globally on May 29, 2026, with Anthropic positioning its latest model as sharper, more honest, and better suited for complex agentic tasks than its predecessor. The release comes just 41 days after Opus 4.7, one of the fastest upgrade cycles the San Francisco-based AI company has run. According to TechCrunch, the quick turnaround may reflect tepid early response to Opus 4.7 from developer communities.
The launch arrives as Anthropic nears the close of a funding round targeting a valuation above $900 billion (roughly Rs 74,700 Cr at current rates). The company also confirmed that Claude Mythos, its advanced cybersecurity-focused large language model (LLM), will reach all customers in coming weeks once safety guardrails are finalised.
StartupFeed Insight
India is already Claude’s second-largest market globally, with run-rate revenue doubling in four months since Anthropic opened its Bengaluru office in February 2026. Founders building on Claude Code now get a model that catches its own errors at four times the rate of Opus 4.7. That reliability gain matters most for Indian startups using AI agents in finance, legal-tech, and SaaS, where unchecked errors carry real business risk. Watch for Indian enterprise adopters, including firms already on the Infosys-Anthropic partnership track, to accelerate their Claude Code deployments. If Mythos-class models arrive by July 2026, Indian IT services firms will face renewed pressure to show clients why they still need human delivery teams at all. By StartupFeed Desk.
What Makes Claude Opus 4.8 Launches Different from Opus 4.7?
The headline improvement is honesty. Anthropic says Opus 4.8 is around four times less likely than Opus 4.7 to let flaws in self-generated code go unreported. Early testers found the model flags uncertainty rather than pushing past it. This closes a criticism levelled at Opus 4.7, which some developers said spent too much time on easy tasks while glossing over hard ones.
Benchmark gains are real but measured. Agentic coding (SWE-Bench Pro) improved from 64.3% to 69.2%. Multidisciplinary reasoning rose from 54.7% to 57.9%. Agentic computer use (OSWorld-verified) moved from 82.8% to 83.4%. Pricing is unchanged from Opus 4.7: $5 per million input tokens and $25 per million output tokens via the API.
Anthropic also launched Dynamic Workflows alongside the model. This research-preview feature lets Claude Code plan large jobs, run hundreds of parallel sub-agents in a single session, and verify outputs before delivery. It is available on Enterprise, Team, and Max plans. A new Effort Control panel lets users on Claude.ai manually set how much compute, and cost, Claude applies to each response.
Claude Opus 4.8 at a Glance
| Metric | Detail | Notes |
|---|---|---|
| Launch Date | May 29, 2026 | Global, same day |
| Pricing | $5 / $25 per million tokens (in/out) | Unchanged from Opus 4.7 |
| Agentic Coding Score | 69.2% (SWE-Bench Pro) | Up from 64.3% in Opus 4.7 |
| Fast Mode Speed | 2.5x faster | 3x cheaper than prior fast mode |
| New Features | Dynamic Workflows, Effort Control | Enterprise and Max plans |
| Misalignment Rate | Near Claude Mythos Preview level | Substantially below Opus 4.7 |
The fast-mode cost cut is the most commercially significant detail. Running Claude at 2.5x the speed at one-third the previous fast-mode price changes the economics for startups that rely on high-volume agentic pipelines.
About Anthropic
Anthropic is an AI safety company founded in 2021 by Dario Amodei and Daniela Amodei, along with several colleagues from OpenAI. Headquartered in San Francisco, it builds the Claude family of LLMs. The company opened a Bengaluru office in February 2026, making India its second-largest market globally by Claude.ai usage. Investors include Amazon, Google, Salesforce, and a range of sovereign wealth and institutional funds. Anthropic’s current annual run rate is reported above $40 billion. Read more about AI and deep-tech news on StartupFeed.
What Is Project Glasswing and Claude Mythos?
Claude Mythos is Anthropic’s most capable model, built with advanced cybersecurity features. It has raised concerns among executives and governments about its potential impact if released without strong safety measures. Under Project Glasswing, a restricted program, Amazon, Microsoft, and Apple are among the firms permitted to use Mythos for cybersecurity work.
“We are making swift progress on developing these safeguards and expect to be able to bring Mythos-class models to all our customers in the coming weeks,” Anthropic said in its May 29 blog post.
Opus 4.8’s misalignment score is now close to the Mythos Preview level, which Anthropic treats as a positive signal. It suggests the safety gap between its broadly available model and its most powerful one is narrowing. For founders interested in what this means for AI strategy, see StartupFeed’s For Founders section.
How Does the Claude Opus 4.8 Launch Affect Indian Startups?
India is already a key market for Anthropic. The company’s India run-rate revenue doubled in the four months following its October 2025 expansion. Software and coding tasks make up 45.2% of all Claude usage in India, the highest share globally, according to Anthropic’s Economic Index data from February 2026.
| Player | India Relevance |
|---|---|
| Anthropic (Claude Opus 4.8) | Bengaluru office; second-largest global market; Infosys partnership for enterprise AI agents |
| OpenAI (GPT-5.5) | Competes on agentic coding; Opus 4.8 outperforms on SWE-Bench Pro |
| Google (Gemini 3.1 Pro) | Strong in India’s developer ecosystem; trails on SWE-Bench Pro vs. Opus 4.8 |
Analysts at Business Standard have noted that Anthropic’s move into direct enterprise service delivery, including a $1.5 billion joint venture with Blackstone, Goldman Sachs, and Hellman and Friedman, puts it in direct competition with Indian IT services firms. Startups using Claude Code gain a more reliable coding partner with Opus 4.8. For a wider read on how AI is reshaping Indian tech investment, visit StartupFeed’s Funding and Deals section.
What Comes Next?
The next milestone is the broad release of Mythos-class models. Anthropic has committed to “coming weeks,” pointing to a likely late-June or July 2026 window. A successful $30 billion-plus funding round at a $900 billion valuation, expected to close imminently, would further accelerate compute investment. Both Anthropic and OpenAI are also said to be exploring IPOs as soon as this autumn. Will Mythos change how Indian enterprises think about cybersecurity budgets?
Frequently Asked Questions
What is Claude Opus 4.8 and when did it launch?
Claude Opus 4.8 is Anthropic’s latest flagship AI model, launched globally on May 29, 2026. The model offers improved performance in agentic coding, multidisciplinary reasoning, and honesty, scoring 69.2% on the SWE-Bench Pro benchmark. Pricing remains at $5 per million input tokens and $25 per million output tokens, unchanged from Opus 4.7.
How does Claude Opus 4.8 launches affect Indian developers and startups?
Indian developers now have access to a more reliable coding model at the same price point. Claude Opus 4.8 launches with a fast mode that is 2.5x faster and 3x cheaper than before. India is Anthropic’s second-largest market globally, and the Infosys-Anthropic enterprise partnership means larger Indian firms also gain access to upgraded agentic capabilities.
When will Claude Mythos be available to all users?
Anthropic confirmed on May 29, 2026 that Mythos-class models will reach all customers in the coming weeks. Claude Mythos is currently limited to select firms under Project Glasswing, a restricted cybersecurity program. Anthropic says it is finalising safety guardrails before a broader release, which analysts expect in late June or July 2026.
