This feed includes only recent releases, roughly from the last 12 months, not the full historical archive. That is why a long-established product may have only a few news items.
1 news
Opus 5 doubles Opus 4.8 score on Frontier-Bench and triples nearest competitor on ARC-AGI 3; pricing unchanged at $5/$25 per million input/output tokens. Weights closed: access only via Claude API, Claude.ai, Claude Code, and Claude Cowork; self-hosting impossible.
Last scheduled beta; RC1 due 2026-08-04, RC2 2026-09-01, final 3.15.0 due 2026-10-01.
The 8B model scored 78.5% on RTEB and 75.5% on MMTEB Retrieval - first place overall. The 1B variant reaches 72.4% on RTEB (27% fewer errors than the previous 1B version), and NVFP4 quantization preserves 99%+ of BF16 accuracy at 2x throughput. Weights and training recipes are open and available on Hugging Face.
LightOn-rerank-LW-2B (2B, a LoRA adapter on Qwen3.5) ranks text passages and document scans with a single model and a single scale - 62.66 NDCG@10 on ViDoRe V3 versus 59.18 for Qwen3-VL-Reranker-2B and 59.40 for jina-reranker-m0. The 4B variant (64.69) beats Qwen3-VL-Reranker-8B (64.23) with half as many parameters. Weights are open on Hugging Face.
A new dashboard with key store metrics directly in Saleor Dashboard, plus an agent-ready data layer for AI agents.
Responsible AI development certificate, previously issued only for base YandexGPT (July 2025), expanded to Alice AI LLM, VLM, and ART. Audit checked development processes, data protection, and error monitoring, not the model itself.
Model (128K context, MMLU-PRO 0.75, MATH500 0.86, Arena-Hard RU 0.69) available in API for individuals only via Freemium mode, weights not published — contrast with 3.5 Ultra opened a week earlier. Standard vendor cloud lock-in for this access tier.
On July 14, 2026, .NET 10.0.10, 9.0.18, and 8.0.29 were released. The scheduled update closed 17 CVEs across all three supported branches (10, 9, 8). Details for each vulnerability are linked to MSRC guidance, without exploit descriptions in the post itself.
Conformer 220M/600M encoders, pre-trained on 2M hours of audio, fine-tuned with CTC decoder on 50K hours for 70+ languages. Best open-source WER for CIS, Kazakh, Kyrgyz, Uzbek; weights open under MIT.
Meta opened first paid OpenAI-compatible API ($1.25/$4.25 per million input/output tokens—roughly one-quarter OpenAI/Anthropic pricing). Model remains closed (weights not published); scores 52 on Artificial Analysis Intelligence Index vs 18 for open Llama 4 Maverick.
Line of three models: Sol — flagship, $5/$30 per million tokens input/output; Terra — GPT-5.5 level at half price, $2.5/$15; Luna — budget tier, $1/$6. Ultra mode deploys sub-agents for complex tasks, claimed support for defensive security (threat modeling, patching, blue team). Weights closed, API/ChatGPT access only.
CDN cache proxy with per-asset and per-thumbnail invalidation via HTTP surrogate keys, plus an SBOM in every release for compliance. Studio: dark theme, Theme Manager, split-screen object translation. The Agent Bundle now also supports documents, including creation and updates, and personalized variants for target groups.
Akeneo Product Cloud gets Agentic Ziggy, AI-native workspace where agents execute PIM tasks from natural-language instructions instead of manual menus. Adds prompt-based AI image edits in DAM (background swap, seasonal variants) and translates retailer syndication errors into plain-language fixes.
Major release: the UI of the setup, monitoring, and development center has been redesigned; the process handling service and message store have been moved to the platform's core services; the internal DatareonDB architecture has been redesigned (indexes, compression); the 1C connector now logs delete events in information registers; and a security audit has been completed.
12 Stitch-based connectors have been added (Circle CI, Shopify, Microsoft Teams, etc.); all Stitch connectors now work not only in Replication, but also in Pipelines and Qlik Open Lakehouse.
Sonnet 5—upgrade to Sonnet 4.6 for agentic coding and tool use, introductory pricing $2/$10 per million tokens through August 31, 2026. Weights closed, access via API/cloud providers only.
Google released Gemini Omni Flash (video generation and conversational editing) and Nano Banana 2 Lite (cheapest image model in the line) in AI Studio, Gemini API and Gemini Enterprise Agent Platform. Both models closed, access cloud/API only.
Workflow- and node-level spans are exported via OTEL to observability systems; most features are also available in the free self-hosted community edition, while custom telemetry tags are enterprise only.
Five agentic capabilities are now generally available: declarative YAML pipelines via coding agents, a conversational catalog/glossary, data quality, and product data agents. Available only as the Qlik Talend Cloud service; there is no self-hosted option.
Microsoft confirmed that the .NET 8 LTS release (36 months of support from November 2023) and the .NET 9 STS release (24 months from November 2024) will both lose security patches and technical support on November 10, 2026. Migration to .NET 10 LTS, supported through November 2028, is recommended.
Four Tier 1 SDKs received beta releases for the new spec: Python mcp v2.0.0b1, TypeScript v2 (split packages), Go v1.7.0-pre.1, C# v2.0.0-preview.1. They implement a stateless architecture, Multi Round-Trip Requests (requesting user input mid-execution without a long-lived connection), routing headers Mcp-Method/Mcp-Name, and standard JSON-RPC error codes. Migration to v2 is optional; existing servers continue to work.
Launched platform for third-party and internal agents within Alice AI chat; first agents from Yandex Taxi and Yandex Lavka, external company access promised later. Tool orchestration follows same path as MCP Hub in AI Studio.
Existing MuleSoft integrations exposed as MCP servers directly from API Catalog — Agentforce discovers/uses them with no custom wiring, governed through Omni Gateway.
Fixed a RocksDB native memory leak in Kafka Streams, along with other critical bugs in the 4.3 branch.
The refund reason can now be specified not only for the entire payment, but also for each return line separately; return reason and refund reason are separated and can be configured independently.
MCP Connector 1.5/1.6: server can pause and request typed missing input (elicitation, opt-in); Call Tool operation gets structured output schemas via inline JSON or external files.
Third-party apps can now add their own widgets directly to the Saleor Dashboard home screen for custom merchant workflows.
Anthropic launched Claude Tag (beta, Enterprise/Team plans), where Claude joins a Slack channel as a regular participant and anyone can assign a task with an @ mention, without a separate bot for the channel. The work model is an agent in the team’s shared channel, not a developer’s personal CLI tool.
Python 3.15.0 beta 3 is out - the feature freeze has passed, and stabilization is underway before release.
A unified product page conversion framework built on MDM/PIM data (core inherited from Riversand): enriched content, reviews, real-time social signals. It targets both human shoppers and AI discovery agents (ChatGPT, marketplace assistants). The claimed figures (+39% add-to-cart, +108% conversion when reading reviews) are vendor-reported and not independently verified.
Zero-touch OAuth via Identity Assertion JWT Grant (ID-JAG): IdP-issued token exchanged for MCP server access token, no per-server consent screen. Adopted at launch by Anthropic, Microsoft, Okta; 7 MCP providers support it (Asana, Atlassian, Canva, Figma, Granola, Linear, Supabase).
Compatibility has been confirmed for the sales management suite with the built-in CRM v3.1 features and the CRM module for 1C:ERP.Holding Management.
Three days after release, a U.S. export directive required Anthropic to disable access to Fable 5 and Mythos 5 for all foreign nationals inside and outside the country. Closed weights in a vendor cloud can be revoked instantly by regulators, leaving enterprises with no lever to preserve access.
Fable 5 — Anthropic's most powerful publicly available model at launch, SOTA on nearly all benchmarks; Mythos 5 — same model without some safety classifiers. Price $10/$50 per million tokens, weights not published: access only via Claude API, Bedrock, Vertex, Foundry — closed weights.
A fix for the 1C subsystem (version 3.0.6.1) has been released to restore stability after platform patch release 3.2.0.5.
1C launched a training program for the mandatory transition to EPD from 01.09.2026; support is being implemented in sales and ERP configurations. A regulatory requirement in CIS.
The information security center added AI code analysis as a control layer for marketplace solutions (2,000+ modules) on top of static rules and expert review. Context-aware analysis catches less obvious classes of vulnerabilities. Vulnerable code blocks publication, and previously published solutions undergo periodic re-audits.
Targeted fixes for critical bugs in the 4.2 branch after Share Groups GA.
On May 27-28, 2026, Saby (formerly the SBIS brand) held an EDI industry forum with case studies on adapting businesses to the mandatory transition to electronic waybills from September 1, 2026.
600+ commits since 4.2.0: broker cordoning, partition size metrics, share group tuning, tiered storage improvements. KIP-1274 marks the classic rebalance protocol as deprecated and slated for removal in Kafka 5.0.
RC was locked on 2026-05-21, and the spec final is 2026-07-28 (a 10-week validation window). The initialize/initialized handshake and the Mcp-Session-Id header were removed, so the server runs behind a standard round-robin load balancer without sticky sessions. Tasks were moved out of core into an optional extension. The new MCP Apps extension provides server-side UI in a sandboxed iframe with JSON-RPC. A formal deprecation policy was introduced (Active → Deprecated → Removed, minimum 12 months); Roots, Sampling, and Logging are marked for removal.
Descriptive UI Layout reads a flow like a recipe in plain English instead of canvas clicks; Multi-App Run Configurations deploys multiple projects on one local runtime through an AI skill; the debugger has been improved (valid JSON payload for pasting into Postman, a separate error scope with exception type/message/cause at the top).
The 17M model beats 33M ms-marco-MiniLM-L12-v2 by +0.051 NDCG@10 (MTEB) with half as many parameters; the 150M model is the strongest mid-tier reranker up to 600M, outperforming the 596M Qwen3-Reranker-0.6B; the 1B model matches the quality of the 1.54B teacher model (a 0.0001 NDCG@10 difference) with 2.4x faster inference. Apache 2.0, open weights.
Validated GS1 data from Syndigo product MDM flows directly into Blue Yonder planning and execution workflows (space planning, network sync), removing manual data cleansing during onboarding. Riversand is not mentioned explicitly, but it is the same multi-domain MDM engine under the hood of Syndigo.
VibeCode has launched and is positioned as the first no-code platform on the CIS market for building business apps with AI. Included: the BitrixGPT 5.5 context agent, Projects AI (task and meeting analysis), Checkin 2.0, and speech analytics 2.0.
500+ fixes; support for PHP 8.5 (PHP 8.2 has been deprecated); Symfony upgraded to 7.4 LTS; support for OpenSearch 3.x and Valkey 8.1/9 added. Laminas MVC replaced with native PHP, Zend_Cache with Symfony Cache, TinyMCE with HugeRTE, and OAuth moved to native PHP functions. Local payment methods added (BLIK, Pay Upon Invoice, ELO), Apple Pay expanded to Chrome/Firefox, Google Pay card vaulting added, and the clearCart mutation introduced in GraphQL.
3.15 hits feature freeze at beta 1; final release scheduled 2026-10-01 per PEP 790.
Elastic Workflows is now GA, combining script-based automation and agentic reasoning for actions based on analysis results (security triage, enrichment, response). Native Prometheus/PromQL support (tech preview) plus TSDB is 2.6x more storage-efficient and 30x faster on queries than Prometheus.
Node 26 includes the Temporal API (Temporal.PlainDate, Temporal.ZonedDateTime) without flags as a replacement for the outdated mutable Date; V8 has been updated to 14.6 (Chromium 146), and Undici to 8.0 (built-in fetch with no API changes). Current release, entering LTS in October 2026.
The node gives the n8n agent an identity in a Microsoft tenant: an email for @mentions in Teams, access to SharePoint sites, and the ability to act as a team member in Microsoft 365 apps. Early preview, available only through the Frontier program.
GPT-5.5 Instant replaced GPT-5.2/5.1 as ChatGPT's default model. Weights closed, access only via OpenAI cloud.
IdP role mapping is now configured inside n8n instead of in an external IdP config; instance bootstrap via environment variables has been added for automated deployment, along with expression-based grouping for complex structures.
Catalog data model now auto-adjusts from retailer rejection logs and search trends; new Flexible AI Sourcing lets enterprises plug own LLM provider (OpenAI, Claude, or other) into enrichment instead of a fixed vendor model. New Upsun-based Extension Platform hosts custom backend apps (ERP integrations, compliance checks) inside Akeneo infra.
Pimcore received its second MQ nomination (the first was the 2025 MQ for DXP). It was named a Niche Player, not a Leader; Gartner highlighted growth areas. The timing matched the release of the Agent SDK beta.
The model ships with automatic filters that block forbidden high-risk cyber requests; it is available natively on Claude Platform and through Bedrock, Vertex AI, and Microsoft Foundry. Self-hosting is unavailable, and gateway routing must keep sessions in the provider's cloud.
DetailPage analyzes digital shelf signals and AI search visibility (Amazon, Walmart, ChatGPT), and attribute/content recommendations are fed back into the managed Syndigo PIM/MDM workflow for syndication and ROI reporting.
Business users (marketers, data stewards) assign tasks to agents in natural language instead of configuration; ready-made agents - Product Creation, Onboarding, Enrichment, Syndication - operate under human oversight. The AI model/provider is not disclosed (closed managed service, not downloadable weights).
Studio (React/TypeScript/Ant Design) reached GA after more than two years of development and completely replaced the legacy ExtJS Admin UI; all code for the classic interface has been removed from the codebase. Bundles now use unified versioning aligned with the platform version.
The Agent SDK is built into the Pimcore core, and agents receive the same permissions, workflows, and governance rules as human users. It was announced at Inspire 2026 together with Data Spine, a unified governance layer for PIM/MDM/DAM/AI.
The new deliveryOptionsCalculate mutation gives the storefront deterministic control over invoking the delivery webhook instead of implicit triggers; PostgreSQL full-text search now supports prefix matching, Boolean operators, and relevance ranking for products, orders, and gift cards.
Meta Superintelligence Labs unveiled Muse Spark: multimodal reasoning model, 1M context tokens, tool use and multi-agent orchestration. Weights not published—end of open-weights as Llama's sole flagship path; Meta claims only 'hope' to open future versions.
A stable unified SDK for .NET and Python for orchestrating single-agent and multi-agent AI workflows (sequential, concurrent, handoff, group chat) with a stable API and an LTS commitment. Built-in connectors include Microsoft Foundry, Azure OpenAI, OpenAI, Anthropic Claude, Amazon Bedrock, Google Gemini, and Ollama; MCP support for calling external tools is already stable, and support for the A2A 1.0 protocol is listed as coming soon.
Official Avalara AvaTax integration adds sales tax calculation at current rates for merchants across markets.
Fine-tuning over November preprint: FP8+MTP at DPO stage halved memory consumption without quality loss, eliminated generation cycles (BPE metric). Ultra and Lightning weights remain open on HuggingFace/GitVerse under MIT.
First agentic PXM platform according to Syndigo, natively built on enterprise multi-domain MDM (Riversand historical core). Agents reason and act across domains (product, supplier, location, compliance, performance) as unified "digital twin", sequentially or in parallel. Only "machine learning + generative AI" disclosed, specific model/vendor unnamed — closed service.
With Yandex Commerce Protocol (YCP), a customer can place an order directly from Yandex Search or Alice without visiting the store website; order data and sales management remain with the seller. Requires a Yandex.Tovary feed and a request through Yandex support.
The patch for runtime 8.0.1.R2025-02-RT fixes CVE-2026-25747 (LevelDB), CVE-2025-22228 (Spring Security), and CVE-2025-33042 (Apache Avro); JobServer has been updated to 8.0.2.20260224. This is routine maintenance, with no new features.
Scheduled security and quality patch for the 2.4.8.x LTS branch (Adobe moved to a monthly patch schedule in January 2026). No new features, only security fixes and bug fixes for the current stable release.
Anypoint Flex Gateway turns existing REST/SOAP APIs into MCP-compatible servers without changing backend service code. General availability follows the governance stage for agent A2A/MCP interactions.
The ModelContextProtocol SDK for .NET has reached v1.0 with full support for the MCP 2025-11-25 specification: Protected Resource Metadata for server authorization, icons for tools/resources/prompts, incremental scope consent based on least privilege, URL-mode elicitation for sensitive data, tool calls inside sampling requests, and experimental Tasks for long-running operations.
4,876 passages from the Judicial College of Victoria Criminal Charge Book plus 100 expert questions were added to the Massive Legal Embedding Benchmark (MLEB, October 2025) - evaluating not only retrieval, but end-to-end answer generation.
Platform update added agents based on third-party DeepSeek-V3.2 (open weights) alongside its closed model — for procurement, incident analysis, code work. Introduced hybrid mode: confidential data on own infrastructure, heavy compute in cloud; agent scenarios cut costs 4+ times.
Anthropic added automatic codebase vulnerability scanning in Claude Code on the web, with ready-made targeted patches for human review (limited research preview, Enterprise/Team). In a pilot with the Government of Alberta, 466 million lines of code were scanned in 20 hours.
Gemini 3.1 Pro released — first minor update with .1 index (Google previously used .5), up to +15% improvement over Gemini 3 Pro Preview on benchmarks, output limit extended to 65,000 tokens. Available via Gemini API/CLI/Antigravity/Vertex AI/Gemini Enterprise, weights closed.
Visual custom metadata, intelligent DataWeave mapping, and project migration from Anypoint Studio directly into Code Builder.
KIP-932 (Share Groups) production-ready: RENEW acknowledgments, adaptive coordinator batching, lag metrics. Streams: server-side rebalance protocol GA for a limited set, DLQ in exception handlers, Java 25 support.
The MuleSoft Connector for MCP has been announced: it makes Mule applications and APIs MCP-compatible without rewriting the backend, with connections to Agentforce, Claude, and other agents through Anypoint Code Builder.
Independent audit confirmed security and reliability controls over observation period, not one-time snapshot (Type 1).
Survey of 1,000 companies: 49% use Bitrix24 as a CRM, and about 20% already use AI tools inside the CRM for routine tasks and handling requests. 69% of users open the system daily.
On MRCR v2 (8-needle, 1M tokens), Opus 4.6 scored 76% vs 18.5% for Sonnet 4.5 — better at retrieving and finding facts in large document sets without explicit RAG index. Model via API/cloud only, weights not published.
Platform's stated direction — action layer for autonomous agents over legacy systems: federated gateway for AI/agent traffic control, LLM token spending tracking, MCP-compatible API out of the box.
Routine bugfix, build and documentation maintenance release for the 3.14 and 3.13 branches.
Starting with Studio 2026-06, the desktop IDE runs only on Java 21; jobs and services themselves remain compatible with Java 17 (Big Data jobs use Java 8). Upgrade through the built-in Feature Manager or the new installer with JDK 21 included.
A patch release that stabilizes the portal, business processes, connectors, and configuration import/export; the detailed changelog is provided in separate documentation.
Akeneo ships a native Model Context Protocol server exposing governed product data access for AI agents, plus a Stripe Agentic Commerce Suite integration so external AI shopping agents can discover catalog items and complete agentic checkout. Also adds Digital Showroom (real-time sales-facing product view) and a custom-component extension type for bespoke UI.
The final 8.5.1 release of the 8.5 line is out, with a new interface (light/dark theme, beta status within the release) and backward compatibility with 8.3. UNF 3.0.13, Retail 3.0.13, and Cashier Workstation 1.1.1 were updated at the same time.
Task runners are enabled by default: Code nodes run in an isolated environment, and access to env variables from Code nodes is blocked. Legacy options have been removed (in-memory binary data), new SQLite pooling delivers up to a 10x boost, and a Publish/Save flow has been added for safe production updates. Breaking changes apply, with a migration report before upgrading; 1.x is supported for another 3 months.
GPT-5.2 (Instant/Thinking/Pro) improved long-context, vision and tool calling; Thinking achieved 80% on SWE-bench Verified and 98.7% on Tau2-bench Telecom. Separate GPT-5.2-Codex — specialized for agentic coding. Weights closed, access only via API/ChatGPT.
Anthropic transferred Model Context Protocol governance to AAIF foundation, co-founders Block and OpenAI, supported by Google, Microsoft, AWS, Cloudflare, Bloomberg. Management and maintainers unchanged — goal to keep protocol vendor-neutral.
Symfony 7 support alongside 6.4, 3-year LTS. Experimental MCP Server support—AI agents/LLMs read Pimcore data via Datahub Simple-Rest. Last release with legacy Admin UI Classic.
Out-of-cycle patch fixing regressions from 3.14.1: exceptions in multiprocessing, exceptions in dataclasses without __init__, segfaults/assertion failures in dict insert path.
Launched public vulnerability research program on Standoff Bug Bounty platform for cloud Bitrix24 — compensation for independent researchers for responsible disclosure. Previously, researcher collaboration was private-only.
Protocol anniversary: experimental Tasks (polling for long-running operations), the Extensions framework for optional modules outside the core spec, OAuth Client ID Metadata Documents for simplified client registration, URL-mode elicitation, sampling with tool-calling support (server-side agent loops). The registry has grown to ~2,000 servers (+407% since September), with 2,900+ contributors in Discord.
Alice AI LLM released in Yandex AI Studio for business; internal testing shows it outperforms DeepSeek V3.1 and Qwen3-235B on 60% of business tasks. Model available only via cloud/API, weights not published.
Opus 4.5 released 6 days after Gemini 3 Pro and 12 days after GPT-5.1, positioned as best for code/agents/computer use at the time. Effort parameter (low/medium/high) at medium delivers Sonnet 4.5 SWE-bench performance with 76% fewer output tokens; 200k/64k context, weights closed.
Launched Nano Banana Pro — image generation based on Gemini 3 Pro, available in Gemini app, AI Studio, Vertex AI, Google Workspace, and Ads. Weights closed, cloud API access only.
New major release: pipe operator |>, built-in URI extension (RFC 3986 + WHATWG), Clone With for readonly properties, attribute #[\NoDiscard], array_first()/array_last(), persistent cURL share handles.
Google released Gemini 3 Pro — reasoning-first model with 1M-token context, precise zero-shot object pointing and records on ARC-AGI-2 (31.1%), AIME 2025 (95%), GPQA Diamond (91.9%), LMArena Elo 1501. Weights closed, access only via Gemini API, Vertex AI and Gemini Enterprise.
Updated Deep Think shows up to 90% on IMO-ProofBench Advanced and gold on written rounds of Physics/Chemistry Olympiad 2025; Deep Think previously won gold on IMO in summer 2025. Mode available via Gemini API/Ultra subscription only, weights closed.
Main feature — extension members: extension of existing extension method syntax to properties, operators, and static members. Added field-backed properties via contextual keyword field, implicit conversions between array/Span/ReadOnlySpan, null-conditional assignment, partial events and constructors, custom compound-assignment operators.
GPT-5.1 Instant/Thinking added adaptive reasoning (model decides whether to reason before responding) and 8 default personality presets. GPT-5.1-Codex-Max and GPT-5.1 Pro released a week later on 19.11.2025. Weights closed, access only via API/ChatGPT; by March 2026 the line was discontinued on ChatGPT in favor of GPT-5.2.
On 11 November 2025, Microsoft released .NET 10 as an LTS release with three-year support until 10 November 2028. Key changes: JIT compiler improvements (inlining, method devirtualization), AVX10.2 support for modern Intel processors, extended post-quantum cryptography (PQC) with Windows CNG support, Aspire 13 for cloud-native development.
Node.js 24.11.0 entered Active LTS on October 28, 2025, with support through April 2028. AsyncLocalStorage now uses AsyncContextFrame by default, URLPattern is globally available, and npm 11 is included.
Product renamed to Alice AI, new generation of models introduced: Alice AI LLM (text), Alice AI Art (images), Alice AI VLM (visual analysis); answers combine text, images, video, and Yandex Maps data. All models closed, access only through Yandex products.
DiskBBQ is a new vector storage format that searches compact clusters directly from disk without loading the index into memory, with sub-20ms latency and low RAM usage. Streams automatically parses unstructured logs and labels fields; Significant Events highlights anomalous patterns against routine noise. Agent Builder is a native chat interface for Elasticsearch data to build custom AI agents.
Current release: the --allow-net flag provides explicit control over network access in the permission model (similar to Deno’s permissions model); Web Storage (localStorage/sessionStorage) is enabled without a flag; V8 has been updated to 14.1 with faster JSON.stringify and built-in Uint8Array base64/hex conversion.
Free-threaded build officially supported (PEP 779), annotation evaluation deferred by default (PEP 649), template string literals / t-strings (PEP 750), experimental JIT compiler in official macOS/Windows binaries, official Android binaries added.
At DevDay 2025, OpenAI released AgentKit — Agent Builder (visual canvas workflows, beta), ChatKit (chat widget, GA), Evals for Agents, and Connector Registry for tool integration. On 3 June 2026, OpenAI announced sunset of Agent Builder and Evals — both leaving platform by 30.11.2026.
Agentic mode in Anypoint for the full application lifecycle: API spec, flow, DataWeave transformation, documentation and MUnit test generation from a prompt. By MuleSoft's internal benchmarks, Mule code quality exceeds direct prompting of general models.
RTEB (beta) combines open and private datasets across 20 languages and domains (law, medicine, code, finance) on NDCG@10 metric to prevent model overfitting to public benchmarks—prior MTEB inflated scores due to public dataset memorization. Used as primary benchmark for 2026 embedding model evaluation (Nemotron 3 Embed and others).
Native VS Code extension (beta) with inline diffs, a checkpoint system that auto-saves the code state before changes (/rewind), subagents for parallel work (backend and frontend at the same time), and background tasks for long-running processes like the dev server. The default model is Sonnet 4.5.
AI features are natively available in every app: "Ask AI search" turns a natural-language question into Odoo filters/groupings, AI agents perform actions on request (quotes, follow-up), and fields are filled automatically by AI. ChatGPT has been added as a provider: the model is cloud-based, there are no weights, and access is only through the OpenAI API.
registry.modelcontextprotocol.io is now open as the central community-driven MCP server registry, API freeze v0.1, with no data-retention guarantees before GA.
KIP-932 adds queue-like consumption on top of partitions (share groups) — mode not yet production-ready, evaluation/testing only.
Flagship model for business: Yandex internal blind testing shows it outperforms GPT-4.1 (OpenAI) in 56% of cases and its predecessor in 58%; hallucinates half as often. Available only via API in the cloud and Alice, weights not published — closed weights.
OpenAI released two reasoning models (117B and 21B parameters, MoE, Apache 2.0 license)—first open weights from the company since 2019. gpt-oss-120b runs on single 80GB GPU, benchmarks near o4-mini, exceeds it on AIME 2024/2025 and HealthBench. Weights published: self-host in your own infrastructure without OpenAI cloud dependency.
MMDiT image-generation model with accurate text rendering across multiple languages, weights open. Enhanced by Qwen-Image-Edit (August 19, 2025) for semantic and point-based editing.
First AI agent from Akeneo — Data Architect Agent — generates catalog data models in days instead of months, claimed up to 70% lower implementation cost. Paired with AI enrichment that extracts attributes and descriptions straight from product images/PDFs, and AI Discovery Optimization scoring how products surface in AI/voice search.
Nothing found for this filter.
