# xpu live

> xpu live is a public comparison directory for AI model API pricing, cloud GPU rental costs, AI agents, humanoid robots, and robot dogs. Its Journal adds practical guides and industry reporting. It brings together provider tariffs, model information, GPU and robot specifications, manufacturer profiles, source links, and dated benchmark results for developers, researchers, and teams.

This file is a curated guide to the website, not a price feed or a complete inventory. Follow the linked pages for current listings, source references, observation dates, availability conditions, and billing units. The catalog covers selected services rather than every offer in the market.

Last reviewed: 2026-10-18. This is a maintained editorial guide; its review date is not the collection time of every linked record.

## Expanded text reference

- [Expanded reference for AI assistants](https://xpu.live/llms-full.txt): Read public catalog rosters, comparison methodologies, selected model/GPU/agent/robot profile extracts, practical Journal guides, and API, MCP and WebMCP documentation in one text file. This is a dated editorial snapshot; use the documented API or MCP service for current observed prices and freshness.

## Start here

- [AI models and rankings](https://xpu.live/): Browse models across text, image, video, audio, and other capabilities; compare listed API prices and sourced intelligence scores. Filters and pagination help navigate the catalog.
- [AI providers](https://xpu.live/providers/): Discover direct developer APIs, cloud platforms, aggregators, and inference providers, with links to their model listings.
- [Cloud GPUs](https://xpu.live/gpu/): Compare GPU configurations, hardware profiles, provider rental rates, and pricing conditions.
- [GPU providers](https://xpu.live/gpu-providers/): Explore GPU rental and serverless compute providers and their collected offers.
- [AI agents](https://xpu.live/agents/): Discover agent projects, their capabilities, interfaces, and official project links.

- [Humanoid robots](https://xpu.live/humanoid-robots/): Explore bipedal platforms, model-specific specifications, price evidence, applications and development stages.
- [Humanoid manufacturers](https://xpu.live/humanoid-manufacturers/): Find the companies behind the listed humanoids and browse their cataloged products.
- [Robot dogs](https://xpu.live/robot-dogs/): Compare dog-like quadrupeds and wheel-legged platforms across inspection, research, education and companionship.
- [Robot dog manufacturers](https://xpu.live/robot-dog-manufacturers/): Browse quadruped makers and their listed products.
- [Journal](https://xpu.live/journal/): Read industry news, model and API updates, compute guides, benchmark interpretation and pricing analysis; search or browse articles by category.
- [Public API documentation](https://xpu.live/api-docs/): Learn how to query the site’s read-only catalog, normalized prices, collection status and recent price changes.

## WebMCP browser tools

- [WebMCP tools and documentation](https://xpu.live/webmcp/): Discover structured browser tools for AI models, AI and GPU providers, agents, humanoid robots, robot dogs, manufacturers and Journal, alongside the GPU explorer tools. Covers page-specific tool names, filters, pagination, public profiles, archive data rules and testing.

WebMCP requires a supporting browser agent with the relevant catalog page open. Tools are discovered on that page; this documentation URL is not a remote MCP endpoint. Loaded catalog listings are distinct from latest archived model and GPU prices. Filter tools change temporary browser UI only; no tool saves catalog data. Browser support is experimental. For access without an open page, use the separate MCP service below.

## MCP access for AI assistants

- [MCP server and connection documentation](https://xpu.live/mcp/): Connect a compatible assistant to the latest public AI model and cloud GPU prices in the archive. Includes setup instructions, tool parameters, response fields and freshness rules.
- [MCP endpoint](https://xpu.live/mcp-server): Public, read-only Model Context Protocol over Streamable HTTP; no API key required. Use an MCP client, rather than opening the endpoint as a normal web page.

Available tools: `search_latest_prices`, `get_latest_price`, `list_price_providers` and `get_price_freshness`. The service returns each provider's latest successful archived collection, with sources, billing units, tariff conditions, observation times and stale flags. It does not provide historical price charts, private account tariffs or administrative logs, and cannot change site data or trigger collections. Read the usage resource `xpu://archive/usage` for comparison rules.

## Featured guide: understand the comparison

- [How we collect & compare](https://xpu.live/methodology/): The essential reading before using a price or ranking. Explains published sources, collection dates, token and GPU billing units, conditional tariffs, missing values, and the limits of headline price comparisons.
- [About xpu live](https://xpu.live/about/): The site's purpose, catalog scope, sources, and approach to comparing AI infrastructure.

## Selected AI model pages

These are representative catalog entry points, not a permanent ranking or endorsement. Each model page separates provider pricing from model information. A provider being listed does not necessarily mean a verified price has been collected.

- [Claude Opus 5.5](https://xpu.live/ai-models/anthropic-claude-opus-5-5/): Compare collected provider tariffs, input/output and cache components, source dates, and model specifications. Uncollected provider prices remain explicitly labeled.
- [Claude Sonnet 5.5](https://xpu.live/ai-models/anthropic-claude-sonnet-5-5/): Review the model's listed API providers and available pricing information.
- [GPT-6 Astra](https://xpu.live/ai-models/openai-gpt-6-astra/): Explore model pricing and deployment-specific tariffs; keep batch and cloud deployment conditions separate when comparing costs.
- [GPT-6.1 Sol](https://xpu.live/ai-models/openai-gpt-6-1-sol/): Review the model listing, provider routes, and collected pricing details.
- [Gemini 4 Argon](https://xpu.live/ai-models/google-gemini-4-argon/): Review the catalog's announced pricing and access conditions. Announced or limited-access offers should not be treated as generally available API capacity.
- [Grok 4.7](https://xpu.live/ai-models/x-ai-grok-4-7/): Compare the listed model routes and pricing information across collected providers.
- [GLM 5.3](https://xpu.live/ai-models/z-ai-glm-5-3/): Explore listed inference providers, tariff components, and model information.
- [Kimi K3](https://xpu.live/ai-models/moonshotai-kimi-k3/): Review available provider listings and compare prices under the displayed billing conditions.

- [GPT-6 Sol](https://xpu.live/ai-models/openai-gpt-6-sol/): Review the cataloged model information and collected provider tariffs.
- [GPT-6 Luna](https://xpu.live/ai-models/openai-gpt-6-luna/): Compare provider routes and source-scoped prices and specifications.

## Selected cloud GPUs

These are useful navigation examples from the catalog, not a measured popularity ranking. Compare the exact GPU edition, memory capacity, rental configuration, and price unit; similar names do not imply equivalent hardware or offers.

- [NVIDIA H100 SXM](https://xpu.live/gpu/h100-sxm/): 80 GB Hopper configuration for training and inference; compare SXM rental offers separately from PCIe and NVL editions.
- [NVIDIA H100 PCIe](https://xpu.live/gpu/h100-pcie/): 80 GB Hopper PCIe configuration with its own hardware profile and rental pricing.
- [NVIDIA H200](https://xpu.live/gpu/h200/): The catalog's H200 SXM profile with 141 GB HBM3e; review memory-intensive workload considerations and provider offers.
- [NVIDIA B200](https://xpu.live/gpu/b200/): Blackwell data-center GPU profile covering the 180 GB HGX configuration, with linked cloud rental information.
- [NVIDIA A100 SXM 80GB](https://xpu.live/gpu/a100-sxm-80/): Ampere accelerator profile for training and inference; keep SXM and PCIe offers distinct.
- [NVIDIA A100 PCIe 80GB](https://xpu.live/gpu/a100-pcie-80/): The PCIe edition of the 80 GB A100, with its own comparison page.
- [NVIDIA RTX 4090](https://xpu.live/gpu/rtx-4090/): 24 GB Ada GPU profile and listed rental options for single-GPU workloads.
- [NVIDIA RTX 5090](https://xpu.live/gpu/rtx-5090/): 32 GB Blackwell GeForce profile and provider rental comparisons.
- [NVIDIA L40S](https://xpu.live/gpu/l40s/): 48 GB Ada data-center GPU profile for AI and visual computing; distinct from L40.
- [NVIDIA RTX PRO 6000 Blackwell](https://xpu.live/gpu/rtx-pro-6000/): 96 GB GPU profile; confirm the precise Server, Workstation, or Max-Q edition in each provider offer.

## Provider entry points

- [OpenAI](https://xpu.live/ai-providers/openai/): Direct API provider profile and linked models in the catalog.
- [Anthropic](https://xpu.live/ai-providers/anthropic/): Direct Claude API provider profile and linked model listings.
- [Google Gemini API](https://xpu.live/ai-providers/gemini/): Google developer API listings, distinct from Vertex AI deployments.
- [OpenRouter](https://xpu.live/ai-providers/openrouter/): Aggregated model routes; router catalog prices and conditions can differ from direct-provider offers.
- [Amazon Bedrock](https://xpu.live/ai-providers/aws-bedrock/): Managed model platform listings with region, account, and deployment-dependent availability.
- [fal.ai](https://xpu.live/ai-providers/fal/): Generative media provider profile and linked catalog models; public price coverage may vary.
- [Runpod](https://xpu.live/gpu-providers/runpod/): GPU provider profile; the catalog identifies the deployment and tariff basis of its collected offers.
- [Vast.ai](https://xpu.live/gpu-providers/vast/): GPU marketplace profile; quotes and availability can change as individual offers change.
- [Modal](https://xpu.live/gpu-providers/modal/): Serverless GPU compute profile; GPU task rates are not complete virtual-machine costs.
- [CloudRift](https://xpu.live/gpu-providers/cloudrift/): GPU rental provider profile with on-demand and reserved configurations; quote-only offers are distinct from published rates.

## Selected AI agents

- [Cline](https://xpu.live/agents/cline/): Coding agent profile with links to supported interfaces and official project information.
- [OpenHands](https://xpu.live/agents/openhands/): Software-development agent ecosystem profile and project links.
- [Goose](https://xpu.live/agents/goose/): Agent profile covering tool extensions and development, research, and workflow uses.
- [Gemini CLI](https://xpu.live/agents/gemini-cli/): Terminal-based agent profile with official project references.

## Humanoid robots and their makers

These are representative profiles, not an endorsement or a universal performance leaderboard. Read the source scope and represented generation on each profile.

- [Unitree G1](https://xpu.live/humanoid-robots/unitree-g1/): Review the represented platform, recorded specifications, configuration-dependent price and evidence sources.
- [Boston Dynamics Atlas](https://xpu.live/humanoid-robots/boston-dynamics-atlas/): Explore the industrial humanoid profile and its documented development or deployment scope.
- [Figure 03](https://xpu.live/humanoid-robots/figure-03/): Review the represented generation, specifications and availability evidence.
- [1X NEO](https://xpu.live/humanoid-robots/1x-neo/): Read the home-humanoid profile, early-access scope and offer conditions.
- [Unitree humanoids](https://xpu.live/humanoid-manufacturers/unitree/): Manufacturer profile and linked bipedal products.
- [Figure](https://xpu.live/humanoid-manufacturers/figure/): Manufacturer profile and the humanoid listed in this directory.
- [Humanoid comparison methodology](https://xpu.live/humanoid-methodology/): Explains evidence tiers, generation-specific sources, development stages and robot price labels.

## Robot dogs and their makers

Industrial inspection systems, developer platforms and companion robots serve different purposes. Compare matching configurations and tasks rather than a single headline speed or payload.

- [Boston Dynamics Spot](https://xpu.live/robot-dogs/boston-spot/): Review the quadruped profile, configuration notes and source-backed specifications.
- [Unitree Go2](https://xpu.live/robot-dogs/unitree-go2/): Explore the listed quadruped and edition-specific specifications and purchase-price evidence.
- [ANYbotics ANYmal](https://xpu.live/robot-dogs/anybotics-anymal/): Review the industrial inspection platform and its documented configuration.
- [DEEP Robotics X30](https://xpu.live/robot-dogs/deep-x30/): Explore the quadruped profile, technical sources and listed price conditions.
- [Sony aibo ERS-1000](https://xpu.live/robot-dogs/sony-aibo-ers1000/): Companion robot profile; keep its purpose distinct from industrial and research platforms.
- [Unitree quadrupeds](https://xpu.live/robot-dog-manufacturers/unitree/): Manufacturer profile and linked robot dogs, including separately listed generations and wheel-legged variants.
- [ANYbotics](https://xpu.live/robot-dog-manufacturers/anybotics/): Manufacturer profile and cataloged inspection platforms.
- [Robot dog comparison methodology](https://xpu.live/robot-dog-methodology/): Explains application scope, evidence order, comparable specifications and manually reviewed pricing.

## Journal: selected practical reading

Use the Journal for dated editorial context and the catalog for the latest recorded prices. News articles describe events at their publication date; distinguish an announcement, reported claim, preview and generally available product.

- [AI Pricing: How to Estimate Model and GPU Costs](https://xpu.live/ai-pricing-cost-guide/): A practical guide to billing units, input/output mix, conditional tariffs, GPU time and cost per useful result. Numerical examples are hypothetical, not current quotes.
- [GPU memory sizing guide](https://xpu.live/gpu-memory-sizing-guide/): Read about model memory requirements, serving overhead and the limits of multi-GPU comparisons.
- [API updates and migration guide](https://xpu.live/api-updates-migration-guide/): Guidance for evaluating API changes and planning migrations.
- [AI benchmark practical guide](https://xpu.live/ai-benchmarks-practical-guide/): Interpret benchmark evidence in the context of the workload and evaluation conditions.

## Public data access

- [API documentation and examples](https://xpu.live/api-docs/): The public read-only base URL is https://xpu.live/wp-json/xpu/v1/. Documentation covers providers, models, GPUs, prices, status and recent change history, including filters and pagination.

API rates describe saved observations and tariff variants, not guaranteed current capacity or the final invoice. Preserve decimal amounts, component units, source URLs, observation times and stale/status flags. A recent price-change endpoint is not necessarily a complete hourly historical series. Use only documented public endpoints; administrative monitoring and private archive access are outside this guide.

## How to interpret robot data

- Robot catalog positions reflect xpu live editorial evidence tiers and name order, not measured intelligence, popularity or independent task-performance scores. NIST references provide comparison context; they do not rank, certify or endorse these catalog entries.
- Humanoid and quadruped methodologies differ. Read the relevant methodology before interpreting the row order, product stage or an earlier generation.
- Robot prices and specifications are manually curated snapshots with source, configuration, currency, market and review date. They are not part of the hourly model/GPU price feed.
- Official, Reported and Reseller labels describe price evidence. A reseller listing does not establish manufacturer authorization; a reported figure is not an official quote. From, Past and Limited labels retain their stated scope.
- Do not interpret deposits, subscriptions, future cost targets or another generation’s price as the listed robot’s purchase price. Contact for pricing and Not published are unknowns, not zero values.
- Robot price sorting groups currencies without exchange-rate conversion. Shipping, tax, accessories, support and integration may change the total. Confirm the represented edition and current terms with the supplier.

## Interpreting model and GPU prices, availability, and rankings

- Preserve the displayed currency and unit: per million tokens, image, second, request, GPU-hour, and other units are not interchangeable.
- Input, output, cache read, and cache write prices are separate components. A lowest-input offer includes output pricing from the same tariff; it is not necessarily the lowest total cost for a workload.
- Standard, batch, priority, long-context, reserved, and deployment-specific offers can have different eligibility and billing conditions.
- GPU prices may exclude CPU, RAM, storage, network traffic, taxes, or other charges. Per-GPU rates are not automatically whole-instance prices.
- Image, video, and audio prices may depend on resolution, duration, quality, or other settings. Compare matching configurations.
- Automated model and GPU price collections are scheduled hourly, but source failures and publication delays can occur. A displayed checking time is an observation time, not a continuous live quote or an availability guarantee. Read the date and status on the linked page.
- Unknown or uncollected prices are not zero and should not be inferred. A provider listing alone is not proof of usable account access, current capacity, or collected pricing.
- Model rankings cite Artificial Analysis where shown. Scores describe a dated evaluation configuration and are not a universal quality judgment. Benchmark settings and priced API tariffs can differ; models without verified scores are not assumed to be weaker.
- Verify the official source linked by the catalog before purchasing, integrating, or relying on a tariff. The directory is not the provider's billing or support system.

## Policies and corrections

- [Commercial policy](https://xpu.live/commercial-policy/): Explains listings, pricing conditions, result ordering, and how commercial placements or referral relationships should be identified.
- [Report an error](https://xpu.live/report-error/): Guidance for reporting incorrect catalog entries with the affected page, published evidence, billing unit, and relevant configuration.
- [Contact xpu live](https://xpu.live/contact/): Questions, catalog feedback, and corrections; public contact email is info@xpu.live. Account and billing issues belong with the relevant service provider.

## Optional

- [Sitemap index](https://xpu.live/sitemap_index.xml): Discover the broader set of published pages beyond the examples selected here.
- [AI model sitemap](https://xpu.live/ta_model-sitemap.xml): Discover additional model detail URLs.
- [GPU sitemap](https://xpu.live/ta_gpu-sitemap.xml): Discover additional GPU profiles.
- [AI provider sitemap](https://xpu.live/ta_provider-sitemap.xml): Discover additional AI provider profiles.
- [GPU provider sitemap](https://xpu.live/ta_gpu_provider-sitemap.xml): Discover additional GPU provider profiles.

- [Journal article sitemap](https://xpu.live/post-sitemap.xml): Discover published articles beyond the selected guides.
- [AI agent sitemap](https://xpu.live/ta_agent-sitemap.xml): Discover cataloged AI agents.
- [Humanoid robot sitemap](https://xpu.live/xpu_robot-sitemap.xml): Discover humanoid profiles.
- [Humanoid manufacturer sitemap](https://xpu.live/xpu_robot_maker-sitemap.xml): Discover humanoid makers.
- [Robot dog sitemap](https://xpu.live/xpu_robot_dog-sitemap.xml): Discover quadruped profiles.
- [Robot dog manufacturer sitemap](https://xpu.live/xpu_dog_maker-sitemap.xml): Discover quadruped makers.

This guide was reviewed against the public website on 2026-10-11. Prices, provider coverage, model availability, and rankings are intentionally not copied as fixed values; consult the linked pages for updated information.