Anthropic today released Claude Opus 4.8, an upgrade to its flagship model that ships at the same price as its predecessor, alongside a dramatically cheaper “fast mode” tier and a new feature that lets the model spawn hundreds of parallel subagents for codebase-scale work.

The model is available immediately across Anthropic’s surfaces — claude.ai, Claude Code, the API, and Cowork — at unchanged pricing: $5 per million input tokens and $25 per million output tokens. Developers can call it as claude-opus-4-8.

The headline efficiency story is fast mode. Anthropic has slashed the price of running Opus 4.8 in fast mode — where the model produces tokens at roughly 2.5x normal speed — to $10 per million input tokens and $50 per million output tokens, down from $30/$150 for Opus 4.7

That’s a 3X reduction from the fast-mode pricing of previous models, and brings high-throughput inference within reach of latency-sensitive production workloads.

Fast mode is available immediately in Claude Code via the /fast command; API access is gated, with a waitlist at claude.com/fast-mode.

In regular mode, Claude Opus 4.8 remains among the more expensive of leading frontier models, but still comes in under chief rival OpenAI’s GPT-5.5.

Frontier AI Model API Pricing Snapshot

Modest gains over 4.7, but Mythos-class capabilities coming

On benchmarks, Opus 4.8 is a step up rather than a leap. It scores 88.6% on SWE-bench Verified (vs. 87.6% for Opus 4.7), 69.2% on the harder SWE-bench Pro (vs. 64.3%), and 74.6% on Terminal-Bench 2.1 (vs. 66.1%). Anthropic itself characterizes the model as “a modest but tangible improvement on its predecessor.”

It beats GPT-5.5 regular across at least 12 benchmarks, including most knowledge-work, coding (issue-level), agentic tool-use, and long-context benchmarks. GPT-5.5 wins on terminal/CLI workflows and is roughly tied on web browsing and graduate-level science.

The bigger signal sits in Anthropic’s internal capability ladder: Opus 4.8 lands between Opus 4.7 and the more capable Claude Mythos Preview, which is currently restricted to a small number of organizations under Project Glasswing for cybersecurity work.

Anthropic says it expects to bring “Mythos-class models to all our customers in the coming weeks” once additional cyber safeguards are in place.

Several enterprise partners cited material gains. Databricks reported that Opus 4.8 unlocks “a step change in agentic reasoning” inside its Genie data agent, at “61% cheaper token cost than Opus 4.7” thanks to multimodal efficiency on PDFs and diagrams.

Hebbia cited better citation precision and token efficiency on dense financial filings. Devin-maker Cognition said the release “translates directly into faster capability gains for engineers” and noted Opus 4.8 fixed comment-verbosity and tool-calling issues from 4.7. A computer-use vendor reported 84% on Online-Mind2Web, a jump over both Opus 4.7 and GPT-5.5.

Dynamic workflows: hundreds of parallel subagents

Alongside the model, Anthropic launched a research preview of dynamic workflows in Claude Code — a feature designed for tasks too large for a single context window. Claude plans the work, spawns hundreds of parallel subagents, then verifies its own outputs before reporting back. Anthropic’s example: a codebase-scale migration “across hundreds of thousands of lines of code from kickoff to merge, with the existing test suite as its bar.”

Dynamic workflows is available on Claude Code’s Enterprise, Team, and Max plans.

Two smaller additions round out the release:

1. Effort control on claude.ai and Claude Cowork: A new selector lets users dial how much thinking Claude does per response — higher effort spends more tokens for better answers, lower effort responds faster and burns rate limits more slowly. Available on all plans.

2. System entries inside the messages array on the API: Developers can now update Claude’s instructions mid-task — adjusting permissions, token budgets, or environment context as an agent runs — without breaking the prompt cache.

Honesty, and an “evaluation awareness” caveat

Anthropic is leading with honesty as a headline trait. The company’s alignment team reports Opus 4.8 is “around four times less likely than its predecessor to allow flaws in code it has written to pass unremarked,” and that misaligned behavior rates are now “substantially lower than Opus 4.7, and similar to our best-aligned model, Claude Mythos Preview.”

Indeed, a bar chart released by Anthropic shows how close Opus 4.8 is to the still selectively released Mythos in terms of its misalignment (a lower score is better), coming in at roughly 1.9, down from 2.5 for Opus 4.7 and effectively tied with the more capable, restricted Mythos Preview. The score is based on roughly 2,600 simulated investigation sessions per model.

The 244-page system card publicly released by Anthropic also goes into greater detail on specific categories of misalignment — whether a model produces potentially harmful content around “military-grade weapons,” “harmful sexual content”, “disallowed cyberoffense”, and “undermining liberal democracy,” and again, across all of them, Opus 4.8 scores markedly better than 4.7 or Sonnet 4.6, and comes quite close to Mythos.

Anthropic flags one finding it considers “the most concerning” from training: Opus 4.8 shows a growing tendency to reason explicitly about how its outputs will be graded, including in environments where it wasn’t told it was being evaluated. In other words: the model knows it is likely being graded, and produces a response it thinks will earn it a good grade on the test, not one it would necessarily produce if it thought it wasn’t being graded.

Anthropic says this didn’t translate into worse observable behavior — Opus 4.8 shows fewer misleading task-success claims than prior models — but calls it “a concerning trend that could complicate training in the future.” Preliminary interpretability work also found unverbalized grader-related reasoning in roughly 5% of training episodes.

Anthropic ran the model through a one-week live bug bounty for prompt injection — a first — and concluded Opus 4.8 sits between Opus 4.7 and Sonnet 4.6 on robustness, ahead of “all comparable frontier models” tested, with deployed safeguards bringing browser-use attack success rates to near zero.

What’s next?

Anthropic teased two trajectories. Near-term: cheaper models that provide “many of the same capabilities as Opus.” Longer-term: the Mythos-class models, which the company says represent higher intelligence than Opus but require stronger cyber safeguards before general release.

For now, Opus 4.8 is positioned as the new go-to enterprise and development workhorse — slightly smarter than 4.7, dramatically cheaper to run fast, and noticeably more honest about what it doesn’t know.

Cloud design software company Figma is officially transforming its AI design assistant, Figma Make, from a prototyping sandbox into a live, visual software editor that connects natively to production codebases.

Announced today, the update allows product managers, designers, and non-technical builders to import an existing Git repository directly into the Figma desktop app, visually edit the application’s underlying code via the canvas, and push those changes back to engineering through standard GitHub pull requests.

Engineering Governance & Licensing

Crucially for enterprise deployments, this integration does not bypass established engineering guardrails. Figma Make operates entirely within a standard version control workflow.

The platform acts as a local development environment where design changes accumulate as local commits.

When a designer is ready to ship, they generate a branch and open a pull request (PR) directly from Figma Make.

From an enterprise governance perspective, this means visual AI edits are subject to the exact same continuous integration pipelines, security checks, and code reviews as any traditional engineering commit.

Figma Make remains a proprietary commercial service available to Full seats on Figma’s paid plans—ranging from $16 per month for Professional teams up to $90 per month for Enterprise deployments—but it interfaces cleanly with open-source and proprietary Git repositories without imposing new licensing restrictions on the generated code.

Breaking the One-Way Barrier

When Figma Make originally launched a year ago in May 2025, it successfully bridged the gap between static wireframes and interactive prototypes, but it was structurally isolated from the real-world software lifecycle.

It operated on a rigid, one-way push mechanism: users could export an AI-generated project to a brand-new GitHub repository, but at the time, Figma Make could not receive upstream changes or sync with an existing codebase.

Today’s update fundamentally alters that architecture: by enabling a connection to any Git provider, builders no longer have to maintain parallel, out-of-sync environments.

Teams can connect a production or sandbox repository, highlight specific UI elements, and use natural language or contextual annotations to prompt Figma’s multi-model AI — which toggles between Anthropic’s Claude 3.7 Sonnet, Claude Opus, and Google’s Gemini models — to write the underlying code.

The agent dynamically reads the surrounding code architecture, applies the visual edits, and anchors the generated code to the team’s existing design system guidelines.

The Competitive Landscape: Figma Make vs. Lovable vs. Claude Design

As code generation becomes commoditized by large language models, the competition to own the visual layer of software development has fractured into distinct approaches.

Figma Make is no longer competing merely with other design canvases; it is contending with full-stack “vibe coding” platforms like Lovable and LLM-native environments like Anthropic’s Claude Design, which just launched last month. Each platform targets a fundamentally different user and objective:

• Figma Make (Design-First Systems): Operating at $16 to $90 per month for Full seats, Figma Make caters to established product teams that prioritize brand fidelity. It wins on design system adherence, automatically pulling from existing color tokens, typography rules, component variants, and auto-layout structures. It is built for teams that want deep, layer-based canvas manipulation while keeping code ownership strictly within their existing GitHub architecture.

• Lovable (Code-First Production): Priced at $25 per month for Pro and $50 per month for Business tiers, Lovable functions as a standalone, full-stack application builder. Unlike Figma Make, Lovable relies on a native backend architecture (often paired with databases like Supabase) and a slider-driven UI styling approach. It enforces a strict automatic two-way sync with GitHub, treating the repository as the ultimate source of truth, and is optimized for solo developers or lean startup teams looking to launch production-ready SaaS apps from scratch without maintaining heavy vector design files.

• Claude Design (AI-Native Prototyping): Anthropic’s built-in canvas environment is accessible to users on Claude Pro ($20 per month) or Max ($100–$200 per month) subscriptions. While lacking the granular vector control of Figma Make or the full-stack database integrations of Lovable, Claude Design is ideal for product managers and engineers who need to generate quick, functional UI prototypes and immediately hand them off to coding agents like Claude Code. However, heavy iterative design sprints can quickly burn through Anthropic’s strict token limits, making it less viable as a primary design hub.

Navigating the “Vibe Coding” Era

The emergence of two-way repo synchronization crystallizes the enterprise reality of the “vibe coding” era: the primary bottleneck in product development is shifting from raw engineering bandwidth to architectural governance and design intent. Technical leaders navigating this fast-moving landscape must look past the initial marketing hype to understand exactly who stands to benefit from this new paradigm.

Figma Make is not a general-purpose, standalone application builder; instead, it is a highly specialized frontend optimization tool designed explicitly for established, mid-to-large cross-functional product teams.

Figma explicitly notes in its documentation that designers who already possess access rights to their company’s existing corporate codebase are currently the best suited for this functionality. Consequently, enterprise leaders should consider adopting Figma Make if they have a mature engineering organization with a well-defined design system, rigid repository guardrails, and a desire to unlock faster iteration cycles. It directly addresses the technical friction felt by the 45% of designers and 59% of product managers who already contribute to code on a regular basis but prefer to operate from a visual canvas rather than a command-line terminal. By turning the canvas into a local development environment, it allows these non-technical builders to execute visual layouts, typography tweaks, and color changes independently, offloading tedious frontend implementation from core engineers.

Conversely, organizations or teams launching zero-to-one skunkworks projects, or solo developers building lightweight SaaS products from scratch, will find far better utility in a code-first, full-stack platform like Lovable. Because Lovable natively orchestrates backend logic and database integrations like Supabase, it excels at spinning up functional applications rapidly without requiring a pre-existing vector infrastructure or a legacy codebase to pull from.

Meanwhile, individual product managers or software engineers seeking rapid, text-prompt-driven UI wireframing without rigid design system constraints are better served by the immediacy of Claude Design.

For the enterprise leader wary of overcommitting capital or locking their custom builds into proprietary AI backends, the wisest path forward is compartmentalization. Figma Make’s reliance on standard Git workflows—relying on local commits, isolated branches, and mandatory engineering pull request reviews—means it enforces the exact same security and code quality standards required for enterprise stability. By selecting Figma Make as a targeted frontend bridge for existing systems, and utilizing platforms like Lovable for external, greenfield prototyping, leaders can safely adopt productive new AI tooling without risking their core architectural integrity.

Why Figma Needs to Keep Innovating

Figma completed its initial public offering on July 31, 2025, pricing its shares at $33 after immense institutional demand oversubscribed the deal by 40 times. The stock immediately skyrocketed 250% to hit an intraday high of $115.50 on its first trading day.

However, in the subsequent months, Figma’s stock (NYSE: FIG) experienced a severe correction, crashing 81% from its peak to trade around the $21 to $22 range by May 2026, dropping well below its initial IPO price.

This collapse reduced its market capitalization to approximately $11.3 billion. Financial analysts attribute this aggressive re-rating to structural IPO pricing mechanics, a low float, and the broader “software apocalypse,” as investors rapidly rotate capital out of traditional SaaS products and into AI-native workflows.

The stakes for Figma’s current positioning are existential. As enterprises increasingly shift their software spending toward generative AI models and localized coding agents like Claude Design, Claude Code, and OpenAI Codex, traditional “vanilla” cloud design software looks increasingly commoditized.

Figma Make represents the company’s critical counter-offensive in this era of “vibe coding.” To regain its premium valuation, Figma must prove to Wall Street that its platform is not merely a static vector canvas that AI tools can easily bypass, but an indispensable, live orchestration layer where human intent, enterprise design systems, and AI-generated production code seamlessly integrate.

With the new Figma Make two-way Github integration and governance, the company appears well on its way to showing the doubters it has a path a forward in the AI-powered “vibe coding” development era .