Every pricing change, model release, deprecation, and incident across the AI/ML stack — detected and published by Change Radar.
Anthropic added an Apr 10, 2026 incident titled “Degraded Performance on Vaults” to the status page. The timeline shows Investigating → Monitoring → Resolved with a fix implemented and monitoring in place. The entry indicates degraded Vault performance impacting that product surface during the incident window.
The main Claude pricing page has been expanded with more detailed billing and pricing information: it now lists payment options (credit card, ACH, invoicing/net terms), billing cycle notes, seat/Team/Enterprise pricing mechanics (seats + API-rate usage for Enterprise), references to model/token pricing and service tiers (priority/standard/batch), and an explicit statement that "Price and plans are subject to change at Anthropic's discretion." The page also contains clarifying language that subscriptions do not cover heavy continuous usage from third‑party agent frameworks and points users toward extra-usage bundles or API keys. These additions consolidate billing, usage-tier, and model-pricing details previously scattered across pages.
OpenAI published an incident entry on its status page reporting elevated login errors for impacted services and that engineers were investigating and implementing mitigations. The incident entry indicates active investigation and mitigation steps.
Multiple threads on Google’s official Gemini API developer forum report widespread 503 (Service Unavailable) and 429 (Too Many Requests) errors across Gemini models and endpoints on Apr 9, 2026. Users report batch jobs stuck, interactions remaining in_progress, and elevated error rates; Google forum activity indicates an ongoing service-impact investigation or at least high-impact user disruption.
Multiple news outlets and user reports indicate Anthropic’s Claude experienced a service disruption on April 9, 2026, with spikes in error reports, login failures, and degraded responses. Downdetector and social posts corroborate a notable increase in user complaints the same day; Anthropic’s public status page did not show an authoritative incident capture in the pages reviewed, but the breadth of independent reports suggests a real user-impacting outage.
Anthropic reported and publicly recorded elevated error rates and connector/login issues affecting Claude on April 9, 2026; the incident shows investigating/monitoring/resolved updates on the public status entries. Independent outlets and outage-tracking sites also reported spikes in user error reports and login/API failures the same day.
Vercel announced an 'Agentic Infrastructure' blog post describing AI SDK 6 which adds an agent abstraction — letting developers define reusable agents across interfaces and workflows and introducing new agent-focused primitives across the platform.
ChatGPT release notes published April 9, 2026 announce a new $100/month Pro plan tier and note updates to how Codex usage works across Plus and Pro plans. The release notes describe plan packaging and usage behavior changes (billing/entitlement and how Codex usage is counted or throttled across plans) but do not indicate immediate API surface changes.
OpenAI ChatGPT release notes (published April 9, 2026) announce a new ChatGPT Pro $100/month plan tier and describe updates to how Codex usage is applied across plans. The notes focus on packaging/billing and usage behavior in ChatGPT (entitlements and quotas) rather than immediate API surface changes.
OpenAI introduced a new $100/month Pro subscription tier for ChatGPT, positioned between the $20 Plus and $200 Pro plans. The $100 plan offers 5x higher usage limits than Plus and 10x Codex usage over Plus (limited time), while the existing $200 Pro plan now provides 20x Codex usage compared to Plus.
Anthropic published a status incident titled “Sonnet 4.6 elevated rate of errors” on their status page with Investigating → Identified updates on Apr 8, 2026. The incident notes elevated error rates impacting Sonnet 4.6 and related Claude services and lists partial outage / degraded performance for some services while a fix is implemented.
Anthropic's status page shows an incident titled "Sonnet 4.6 elevated rate of errors" with Investigating/Identified updates on 2026-04-08 (timestamps show investigation beginning at 06:23–07:06 UTC and an Identified entry at 07:59 UTC). The incident was active on the status page at the time of collection and was listed among recent/past incidents for April 8, 2026.
OpenAI posted an incident titled "Elevated Errors with Login" on its status page for 2026-04-08. The incident timeline shows an Identified update at 17:23 and mitigation/monitoring updates, with a Resolved status at 17:44 on the same day. Affected components listed were ChatGPT and the APIs.
Developers reported multiple issues with Google Gemini models starting April 8, 2026: a forum thread documents that gemini-2.0-flash with google_search grounding returned only triple backticks (```) instead of content, and other posts report 503 / 'No capacity available' errors for high-tier Gemini models. These reports are from Google's AI developer forum and community channels.
Anthropic posted incident history showing elevated errors on Claude.ai and related product surfaces on April 7–8, 2026 (authentication/login errors affecting web, mobile, platform.claude.com, and /login attempts via Claude Code). The April 8 authentication issue was reported as investigated and resolved (timestamps shown on the status page).
OpenAI posted a status incident titled “Elevated Errors with Login” reporting that users experienced login issues affecting API and ChatGPT components. Engineers identified the issue, implemented a mitigation, and the incident was marked resolved with monitoring in place.
Anthropic’s public status page records multiple elevated-error incidents affecting Claude services and workspace creation across Apr 7–8, 2026. The incidents were investigated, mitigations deployed, and the issues were resolved with monitoring status updates posted on the status page.
Vercel published a security advisory (CVE-2026-23869) describing a high-severity React Server Components vulnerability that can lead to excessive CPU usage and denial-of-service when specially crafted requests hit App Router server functions. Vercel deployed WAF mitigations but recommends immediate upgrades to patched framework releases to fully remediate the issue.
Vercel published a security advisory (CVE-2026-23869) describing a high-severity React Server Components vulnerability that can trigger excessive CPU usage in App Router server functions and result in denial-of-service. Vercel deployed WAF mitigations platform-wide but recommends upgrading to patched framework releases for full remediation.
langgraph-cli 0.4.21 and 0.4.20 were published (08 Apr). The CLI gained a validate command and added remote build support for langgraph deploy flows; there are also CI/dependency bumps. These changes add new developer tooling for building and validating graph/agent deployments.
langchain-core 0.3.84 was published on 2026-04-08 and includes additional prompt sanitization changes. The release notes indicate changes to how prompts are sanitized (likely affecting template or prompt normalization) but do not list explicit API removals.
Microsoft updated the Azure OpenAI (Foundry) quotas & limits page to add a new explanatory section describing causes of 429 (Too Many Requests) responses. The note explains that 429s can occur due to input/context length rejections (HTTP 400), evaluation of potential token usage (e.g., max_tokens), or distributed rate-limiting enforcement that may not be precisely reflected in aggregated usage metrics.
Documentation pages for two Gemini service tiers were published/updated: Flex inference (Preview) and Priority inference (Preview). Flex is documented as a lower-cost, sheddable tier (~50% discount vs standard) intended for latency-tolerant background workflows; Priority is documented as a premium, low-latency tier (≈75–100% premium) for business-critical workloads with graceful downgrade behavior.
Anthropic’s public status page shows a new incident on Apr 8, 2026: "Sonnet 4.6 elevated rate of errors." The incident timeline shows Investigating → Identified → Monitoring → Resolved with timestamps on Apr 8 (fix implemented and monitoring at 09:33 UTC). The entry notes Sonnet 4.6 exhibited errors above baseline between ~07:00–09:50 UTC and that a fix was applied.
Multiple community reports on the official Google developers forum describe Gemini Batch API jobs remaining stuck in JOB_STATE_PENDING for extended periods (20+ hours up to several days). Google staff replied in the thread indicating prior fixes and guidance, but reports show the problem has recurred/continued intermittently through early April 2026.
OpenAI status reported a partial outage affecting some users' ability to access the ChatGPT website when using VPNs. The incident was investigated, a mitigation applied, and services were reported as recovered on Apr 7, 2026 at 11:49 UTC.
Anthropic reported a major outage that affected access to the Claude website and Claude Code for many users. The company's status dashboard and media reports indicate a fix was applied and success rates returned to normal; Anthropic stated they are continuing to monitor.
Vercel added a premium 'Fast Mode' for Claude Opus 4.6 on AI Gateway. The changelog shows how to enable it via providerOptions (e.g., anthropic: { speed: 'fast' }) or a settings.json flag (fastMode: true) and notes it is an early/experimental high-speed option with materially higher pricing (fast mode priced at ~6x standard Opus rates).
Vercel added GLM 5.1 (zai/glm-5.1) to the AI Gateway model roster. The changelog and model page show usage examples using the AI SDK (set model to 'zai/glm-5.1') and reference provider capabilities, pricing, and provider metadata. This is a model release that can change routing/selection and must be accounted for in provider allowlists and routing tests.
LangChain core 1.2.27 was released on 2026-04-07. The release notes list a bugfix that correctly handles symlinks in the deprecated prompt save path and a comment explaining the pygments>=2.20.0 bump. The change is a bugfix in prompt saving/loading paths and does not advertise API-breaking changes in the release notes.