Spring AI 2.0.0-M8 Released with Anthropic Rate Limit Headers and Provider Fixes
Spring AI 2.0.0-M8 was released on May 27, 2026, advancing the pre-release train one step closer to the 2.0.0 GA that is part of the June 1–5 Spring release train. The milestone is primarily a stabilization release, but two additions stand out for teams working at production scale with multiple AI providers.
The most developer-relevant addition is Anthropic rate limit header exposure. The Anthropic API returns rate limit information in response headers — tokens used, tokens remaining, and the reset timestamp — but until M8 that data was discarded by the Spring AI integration. M8 surfaces these values through ChatResponseMetadata, giving applications the ability to observe quota consumption in real time, implement backpressure, and surface meaningful errors to end users when limits are near rather than letting them hit a hard 429. Teams running batch inference jobs or high-throughput chat applications will find this particularly useful.
On the compatibility side, M8 fixes a regression in the Google GenAI starter that declared an incorrect transitive dependency on the embedding module, breaking projects that used the starter without a vector store. The PGVector integration received a fix for a startup failure caused by a missing spring-boot-starter-jdbc dependency in certain autoconfiguration paths. An OpenAI authentication regression that incorrectly required an API key even in cookie- or session-authenticated flows was also resolved, relevant for teams running Claude-compatible or self-hosted OpenAI-compatible endpoints.
M8 is available from Maven Central under org.springframework.ai:spring-ai-*:2.0.0-M8. The Spring AI 2.0.0 GA is part of the broader Spring release train currently targeting June 1–5.
Spring Boot 4.1 GA Release Train Opens This Week
The Spring team confirmed on May 11 that the full May release train — originally scheduled for May 11–22 — has been rescheduled to June 1–5, 2026. The train includes Spring Boot 4.1.0 GA and coordinated releases across the Spring portfolio: Spring Security 7.1, Spring Integration 7.1, Spring Modulith 2.1, Spring AMQP 4.1, Spring for Apache Kafka 4.1, and Spring Vault 4.1.
Spring Boot 4.1's headline additions center on observability and cloud-native runtime behaviour. The first release candidate added support for OpenTelemetry Protocol (OTLP) SDK exporter environment variables, allowing operators to configure exporters using the standard OTEL_EXPORTER_OTLP_* environment variable set without any Spring-specific configuration. Support for Spring Framework's LazyConnectionDataSourceProxy landed in the same RC, giving applications that use connection-lazy patterns — common in service-mesh environments with unpredictable startup ordering — first-class Boot auto-configuration.
The coordinated portfolio releases bring their own improvements. Spring Integration 7.1 refactored RedisLockRegistry to use Redis 8.4+ Compare-and-Set and Compare-and-Delete commands for better lock management performance. Spring Modulith 2.1 introduces @ModuleSlicing, an annotation that eliminates integration test failures caused by cross-module bean visibility. Spring AMQP 4.1 adds @EnableAmqp for semantic alignment with @EnableRabbit and a new setStopListenerOnFatal() method on ConditionalRejectingErrorHandler. For teams that depend on any of these projects, the coordinated release means compatible dependency sets are available from day one rather than requiring manual BOM reconciliation.
Spring Boot 4.1.0 GA release notes are published at the Spring Boot GitHub wiki, and the spring-boot-starter-parent BOM will pick up all compatible portfolio versions automatically.
Links & Sources
You May Also Enjoy
