The latest AI news and updates of OpenRouter — Unified API for accessing hundreds of LLMs across providers with intelligent routing and optimization. Covering OpenRouter's latest product updates, launches, and analysis from the past 90 days.
OpenRouter18h agoOpenRouter reports DeepSeek V4.1 Flash processed one trillion tokens in its first 24 hours, on pace for 2.8 trillion in 48 hours. 90% of this volume consisted of cache reads priced at approximately $0.006 per million tokens. This rate is five times cheaper than comparable models like GLM-5.3 Flash.
OpenRouterSep 10OpenRouter launched a beta server tool, openrouter:shell, that lets any model run commands in a hosted Linux container. The accompanying Files API enables moving files into and out of the sandbox. Usage costs $0.0001 per active second, with a 30-second minimum for cold container starts, and supports both OpenAI and Anthropic tool specifications.
OpenRouterSep 9OpenRouter has made US In-Region Routing generally available on Business and Enterprise plans. Requests sent to the US endpoint are decrypted and processed entirely within the region, ensuring end-to-end data residency. This update allows teams to access Chinese open-weight models like DeepSeek V4 Pro and Kimi K3 through US and EU providers without data leaving the specified jurisdiction.
OpenRouterSep 8OpenRouter added Nex AGI's agentic N2.5-Pro and N2.5-Mini models to its unified API. Both models feature tool calling and adjustable reasoning effort for agentic coding tasks. Nex-N2.5-Pro supports text and image inputs for visual feedback loops, while Nex-N2.5-Mini provides a lighter, text-only alternative. Both models are available for free for a limited time.
OpenRouterSep 8OpenRouter launched Inception AI's Mercury 2.5, a diffusion-based reasoning model that generates tokens in parallel. It features a 260,000-token context window, tool calling, and structured output support. The model is available at $0.04 per million input tokens and $0.15 per million output tokens, ranking as the fastest model on the platform that runs on standard hardware.
OpenRouterSep 5OpenRouter added Microsoft's MAI-Image-2.6 and MAI-Image-2.6-Flash models to its unified API. Both support multi-image editing, web grounding, dynamic aspect ratios, and up to 1.5K output. The flagship model costs $5 per million tokens, while the Flash variant provides 2.8x faster generation than GPT-Image-2-Medium at $1.75 per million tokens.
OpenRouterSep 3OpenRouter has added Microsoft AI's MAI-Transcribe 2 to its unified API. The model supports 60 languages with automatic language identification, code switching, speaker diarization, and word-level timestamps. It is priced at $0.10 per hour and ranks #1 on the FLEURS multilingual benchmark.
OpenRouterSep 3OpenRouter launched a video benchmark suite that compares generation quality, cost, and speed across multiple AI video models. The platform provides side-by-side testing on curated prompts, including physics and prompt-adherence challenges. Users can generate custom prompts to evaluate model performance, with specific data showing Seedance 2.5 excels at physics while MiniMax H3 Max offers lower cost and faster generation.
OpenRouterSep 2OpenRouter has added the MiniMax H3 Max video-generation model to its unified API. The model generates 5–15 second clips from text or keyframe images in ~30 seconds for $0.40, significantly faster and cheaper than the previous H3 version. It supports 480p and 768p resolutions across six aspect ratios, with pricing starting at $0.05 per second.
OpenRouterSep 1OpenRouter has added Anthropic’s Claude Fable 5.1 to its unified API, offering a direct upgrade for Fable 5 workloads. The model improves performance in agentic coding, visual generation, and analysis tasks. It maintains existing input and output pricing while reducing prompt cache read costs by 75% to 0.25 dollars per million tokens.
OpenRouterSep 1OpenRouter integrated with Render Workflows to process LLM prompt batches as individual, retriable task runs. This architecture fans out prompts to avoid holding HTTP requests open, allowing Render to provision compute on-demand for each call. The integration supports Node and Python, using OpenRouter’s Auto Router to select models for each prompt while tracking model usage per task.
OpenRouterSep 1OpenRouter exclusively launched Inception AI's Mercury 2.5 Preview, a reasoning model that uses parallel token generation to reach 1,107 tokens per second. It features a 260,000-token context window, tunable reasoning, and parallel tool calling. The model is priced at $0.04 per million input tokens and $0.15 per million output tokens, targeting latency-sensitive production workloads like voice and coding agents.
OpenRouterAug 27OpenRouter added Meta's Muse Image model to its unified API, priced at $0.01 per image. The model supports text-to-image generation and targeted editing, grounded by search to maintain factual and visual references. It features a 65,536-token context window and is available for high-volume creative workflows through the OpenRouter Image API.
OpenRouterAug 27OpenRouter has added Alibaba's Qwen3.8 Flash to its unified API. The multimodal reasoning model supports text, image, and video inputs, featuring a 1-million-token context window and 131,072-token output limit. It is priced at $0.16 per million input tokens and $0.47 per million output tokens, supporting coding assistance, agentic workflows, and visual analysis.
OpenRouterAug 26OpenRouter added Recraft V4 Styles to its unified API, featuring four variants for raster and vector image generation. The models reproduce the texture, composition, color, and rendering of one to ten user-supplied reference images without requiring training. Pricing starts at $0.035 per image plus a $0.005 style-creation fee per request.
OpenRouterAug 26OpenRouter added Makora as a new inference provider, expanding its unified API with four model endpoints. The launch includes DeepSeek V4 Flash, Kimi K3, and two quantization variants of GLM 5.2 in FP8 and NVFP4 formats. Requests can now be routed through Makora to access these models directly via the OpenRouter platform.
OpenRouterAug 25OpenRouter updated its :nitro model variant to admit priority service tier endpoints into the eligible routing pool. These endpoints now compete alongside default endpoints based on measured throughput, winning only when they are genuinely the fastest option. Billing follows the tier served, ensuring priority rates apply only when a priority endpoint successfully handles the request.
OpenRouterAug 25OpenRouter now hosts Alibaba’s Wan 3.0, a video generation model supporting text-to-video, image-to-video, and reference-guided generation. It produces 2–30 second clips at 480p, 720p, or 1080p resolutions. Pricing starts at $0.0425 per second, with a 15% discount currently applied across all resolutions.
OpenRouterAug 22OpenRouter added DeepSeek V4 Flash Vision Exp to its API, an experimental model supporting image input at V4 Flash pricing. The model matches V4 Flash 0731 on text agent benchmarks and outperforms Opus-4.8 on Agents' Last Exam and ZeroBench. It features a 1-million-token context window and supports tool calling for multimodal agent workflows.
OpenRouterAug 22OpenRouter now hosts Black Forest Labs' FLUX Video Upscale, enabling video upscaling to 2K or 4K resolution. The model offers two processing modes: precise for faster results with identity retention, and creative for enhanced detail. It is available via the OpenRouter Video API, with pricing starting at $0.075 per megapixel-second.
OpenRouterAug 22OpenRouter is discounting OpenAI's GPT-5.6 Sol model by 50% across its batch, flex, and priority tiers. The promotion, which runs through September 18, 2026, reduces flex tier pricing to as low as $1.25 per million input tokens and $7.50 per million output tokens. The discount applies automatically to non-BYOK requests routed through the OpenAI provider.
OpenRouterAug 22OpenRouter launched a Cost per Session data layer and API endpoint to measure the real-world cost of using models across agentic tasks. By calculating 30-day medians of per-session USD spend, the metric accounts for token efficiency rather than just per-token pricing. Developers can query this data via the Dataset API or view rankings on the OpenRouter platform.
OpenRouterAug 22OpenRouter added Meta's Muse Spark 1.2 Contributor tier to its API, priced at $0.10 per million input tokens and $0.20 per million output tokens. This reasoning model supports a 1-million-token context window and multimodal inputs. The tier allows Meta to use prompts and outputs for product improvement, offering a lower-cost option for experimentation and early-stage projects.
OpenRouterAug 22OpenRouter upgraded its Activity dashboard and Analytics API to provide granular usage tracking per agent, model, and request. The update exposes metrics including spend, token volume, cache hit rates, and latency. Integration of the openrouter-analytics skill enables autonomous agents to perform cost reviews, identify runaway models, and optimize pipeline spending through direct data queries.
OpenRouterAug 22OpenRouter added Ox Alpha, a free stealth model designed for coding and sustained agentic workflows. It features a 1-million-token context window and supports text, image, and video inputs. The anonymous third-party provider does not use prompts or completions for training. Ox Alpha is available now for programmatic access via the OpenRouter API.
OpenRouterAug 22OpenRouter, the AI model marketplace and gateway processing over 10 trillion tokens daily across 400 models, is joining Stripe. The company will continue operating under its current name and roadmap, maintaining its commitment to neutral, multi-model routing. The acquisition aims to accelerate the development of its infrastructure for the global AI ecosystem.
OpenRouterAug 15OpenRouter launched Ori Grok Build, a CLI harness that runs xAI’s Grok Build tool through OpenRouter’s infrastructure. The tool uses existing OpenRouter credentials and organization guardrails, while automatically disabling xAI telemetry and error reporting. It supports native Grok flags like model selection and reasoning effort, maintaining existing agentic workflows without requiring new API keys.
OpenRouterAug 14OpenRouter launched Ori DeepSeek Harness, a CLI tool that configures the DeepSeek agent framework to run on its platform. The tool enables OAuth sign-in, organization-wide guardrails, and unified billing across 500+ models, allowing the agent to operate without manual API key management or complex configuration changes.
OpenRouterAug 14OpenRouter launched Ori Prime Agent, a CLI harness that runs Prime Intellect’s autonomous agent on OpenRouter’s infrastructure. The tool provides access to over 500 models, replaces manual API keys with OAuth authentication, and applies organization-wide guardrails to every session. Usage and spend from the agent are consolidated into a single OpenRouter bill.
OpenRouterAug 14OpenRouter launched live Web Search Benchmarks to rank search configurations across four task suites, including BrowseComp and HLE. The leaderboards compare models, engines, and search budgets by quality, cost, and speed. Data shows that increasing the search budget from one to 25 turns is the most effective way to improve answer quality for agentic workflows.
OpenRouterAug 14OpenRouter now hosts Voyage Code 4, a code embedding model from Voyage AI by MongoDB. Designed for coding agents, the model supports Matryoshka embeddings at 256, 512, 1024, and 2048 dimensions with multiple quantization options. It features a 32,000 token context window and is available for $0.12 per million tokens.
OpenRouterAug 13OpenRouter is offering an exclusive 50% discount on Google DeepMind's Gemini 3.7 Flash model through August 27, 2026. The promotion reduces pricing to $0.375 per million input tokens and $1.875 per million output tokens. The model supports multimodal inputs and is optimized for fast agentic workflows and complex multi-step reasoning tasks.
OpenRouterAug 12OpenRouter launched Ori Pi, a CLI tool that configures the Pi coding agent to run on OpenRouter with one-install setup. The tool applies organization-wide guardrails and model allowlists to every session. It includes session-level routing toggles for speed or zero-data-retention modes and automatically handles provider API keys to prevent configuration conflicts.
OpenRouterAug 12OpenRouter launched DeepSeek V4 Pro 0813, the general availability release of the mixture-of-experts model. It supports a 1-million-token context window and 384,000-token output limit, priced at $0.435 per million input tokens. DeepSeek reports agent benchmark gains over the preview, including 62.7 on DeepSWE and 83.3 on CyberGym. The model is currently hosted by one provider.
OpenRouterAug 12OpenRouter added xAI's Grok 4.6 to its unified API. The model features a 500,000-token context window and shows performance gains over Grok 4.5, including benchmark leads on GPDVal-AA V2. It remains priced at $2 per million input tokens and $6 per million output tokens.
OpenRouterAug 12OpenRouter updated Ori Eval with five improvements, including pre-run cost estimation via the new --pilot flag and parallel model comparisons that run about 5x faster. The tool now supports headless operation for CI pipelines, provides failure-cause reporting with provider attribution, and warns when an LLM judge is related to a candidate model.
OpenRouterAug 12OpenRouter adds Meta’s Muse Glimmer 30B, a 30-billion-parameter dense multimodal model, to its unified API. The model, licensed under Apache 2.0, supports a 131,072-token context window and is optimized for local agentic workflows. It achieves benchmark scores of 75.5 on MCP Atlas and 51.2 on SWE-Bench Pro, with pricing starting at $0.30 per million input tokens.
OpenRouterAug 12OpenRouter added xAI’s Grok Imagine Image 2.0 to its unified API, supporting text-to-image generation and reference-based editing. The model handles typography, layout, and multi-part visuals with fidelity across photography and design. It is available for programmatic access starting at $0.04 per image, with support for up to three reference photos per request.
OpenRouterAug 10OpenRouter upgraded its Auto Router to select models based on aggregate community spend data from the past seven days. The router classifies prompts into 30 task types to match models to real-world usage. Benchmarks show the new router matches or beats previous performance across five domains, reducing costs at the default tier and improving accuracy at the max tier.
OpenRouterAug 9OpenRouter added ByteDance’s Seedance 2.5 joint audio-video generation model to its API. The model produces 30-second video clips from up to 50 multimodal reference inputs, including images, video, and audio. It supports multi-round extensions and precise visual or audio editing, allowing for consistent long-form storytelling through programmatic access.
OpenRouterAug 8OpenRouter reports that GPT-5.6 Luna token volume surged over 10x following a 10x price reduction on its platform. This usage spike has pushed the model past GLM 5.2 in total platform volume. The trend illustrates the Jevons paradox, where lower inference costs drive a disproportionate increase in demand for model usage.
OpenRouterAug 8OpenRouter now integrates with Netlify Agent Runners and AI Gateway, providing access to hundreds of open and commercial models. This integration supports models like DeepSeek, Qwen, GLM, and Kimi directly within Netlify production infrastructure. The platform maintains consistent deployment pipelines while offering model flexibility for software creation.
OpenRouterAug 6OpenRouter launched two new task-specific leaderboards tracking model usage by real spend share. DeepSeek V4 Pro currently leads the shell execution category, while Moonshot AI’s Kimi K3 holds the top position for tool dispatch. These rankings provide visibility into which models developers prioritize for specific agentic workflows based on actual platform spending data.
OpenRouterAug 5OpenRouter added Meta’s Muse Spark 1.2 reasoning model to its API, priced at $1.25 per million input tokens and $4.25 per million output tokens. The model supports a 1-million-token context window and is optimized for multi-file refactors, long debugging sessions, and multi-step coding agent tasks. Global access to both Muse Spark models is now available.
OpenRouterAug 5OpenRouter released Agent SDKs for Python and Go, providing language-specific toolkits for building AI agents. Both SDKs are automatically synchronized with the TypeScript Agent SDK, ensuring identical API surfaces and feature parity across all three languages. The SDKs support model calls, tool definitions, and multi-step agent loops through the OpenRouter Responses API.
OpenRouterAug 4OpenRouter launched Ori Harness, a CLI tool that configures coding agents like Claude Code, Codex, OpenCode, and Hermes to use OpenRouter with optimized settings. For Claude Code, the tool automatically enables configurations that nearly halve the system prompt and reduce task costs. It also routes Claude Opus requests across eight endpoints to maintain uptime without manual retry logic.
OpenRouterAug 4OpenRouter launched interactive image model playgrounds for generating, editing, and combining images, starting with OpenAI's GPT Image 2. The browser-based interface provides a visual sandbox for testing text-to-image generation, image editing, and image combination without writing API code. These playgrounds offer a direct way to evaluate model capabilities and performance.
OpenRouterAug 4OpenRouter added lifecycle hooks and async tools to its Agent SDK. Lifecycle hooks intercept agent loop events to block destructive commands, redact secrets, or track usage. Async tools support background execution and deferred resolution, allowing agents to pause for human review or external webhooks without blocking the conversation.
OpenRouterAug 4OpenRouter added Alibaba's Qwen3.8 Max, a 2.4-trillion-parameter flagship (95B active parameters) for long-horizon coding, research, and multimodal agent work. Open weights are due next week — a first for a Qwen Max-class model. Qwen reports gains over Qwen3.7 Max across long-horizon and visual agent benchmarks: PaperBench 64.8→93.0, OSWorld-Verified 73.3→86.1, Vision2Web 42.1→69.0.
OpenRouterAug 3OpenRouter launched Ori Eval, an agent that explores codebases to write and run model evaluations. It assesses performance criteria, tests multiple models via the OpenRouter API, and provides a ranked comparison. Ori Eval integrates into CI pipelines to block deploys on regressions and converts bug reports into permanent test assertions.
OpenRouterJul 31OpenRouter adds Runway’s Aleph 2.0 and Gen-4.5 models to its unified API. Aleph 2.0 supports text-guided video editing and keyframe-based adjustments, while Gen-4.5 provides cinematic text-to-video and image-to-video generation. Both models are available for programmatic access, with Aleph 2.0 priced at $0.28 per second and Gen-4.5 at $0.12 per second.
OpenRouterJul 31OpenRouter added MiniMax’s open-weights H3 model to its unified API, supporting multimodal inputs including text, image, audio, and video. The model generates native audio-visual output and handles instruction-led edits, text rendering, and video-to-video motion transfer. It is available for programmatic access at $0.13 per second of video output.
OpenRouterJul 30OpenRouter added six Voyage AI by MongoDB models to its unified API, including three Voyage 4 text embedding models, a multimodal embedding model, and two instruction-following rerankers. All models support a 32K context window. The Voyage 4 series features a shared embedding space, allowing interoperability between models for indexing and querying tasks.
OpenRouterJul 29OpenRouter added Alibaba’s Qwen3.7-Flash, a vision-capable reasoning model designed for multimodal agents and visual coding. The model supports a 1-million-token context window and tool use, with pricing at $0.03 per million input tokens and $0.13 per million output tokens. It is now available for programmatic access through the OpenRouter API.
OpenRouterJul 27OpenRouter added Moonshot AI’s Kimi K3 multimodal reasoning model to its unified API, offering access through multiple third-party providers. The platform supports reasoning tokens for step-by-step thinking and provides performance metrics for provider selection. Kimi K3 Fast variants from providers including wafer_ai and Fireworks are scheduled for release soon.
OpenRouterJul 27OpenRouter, in partnership with OpenAI, is offering a 50% discount on GPT-5.6 Terra and Luna models. The price reduction applies to input, output, and cache read costs for the first-party OpenAI provider. Pro versions of both models are included in the limited-time offer, which is available exclusively through the OpenRouter API.
OpenRouterJul 25OpenRouter launched a Discover page to explore AI models, featuring live rankings and performance metrics. The page highlights Anthropic’s Opus 5 and OpenAI’s GPT 5.6 Sol as top performers, while providing access to routing solutions and always-latest model aliases. Filters surface models by modality, speed, and cost-efficiency to support model selection.
OpenRouterJul 24OpenRouter added xAI’s Grok STT 1.0 to its unified API. The model supports transcription in 25 languages, word-level timestamps, and speaker diarization. It is available for $0.10 per audio hour.
OpenRouterJul 24OpenRouter added Ant Group’s Ling-3.0-flash, a 124B-parameter Mixture-of-Experts model with 5.1B active parameters per token. Designed for token-efficient agentic inference, the model features a 262,144-token context window and 32,768-token output limit. It is available for free through August 3, 2026.
OpenRouterJul 23OpenRouter now provides access to Alibaba’s Qwen-Audio-3.0-TTS Flash and Plus models through its unified API. These text-to-speech models support 16 languages and accept natural-language direction, including expressive inline cues like [gasp] or [angry]. Flash is optimized for low-latency real-time interaction, while Plus prioritizes voice fidelity and naturalness for polished speech generation.
Frequently asked questions