All projects matching “model”

x402engine — Meta's open-source model — great general-purpose performance at very low cost

SWARM Read Page - Retrieve the content of a URL as clean, LLM-ready text (headings with #, lists with -; menus, scripts, styles and markup removed). Use when an agent needs to read a web page, article or docs page without paying model tokens for raw HTML: on 15 public pages we measured, the text was a median 97% smaller than the HTML (about 67,000 fewer tokens per page). Requires `url`; HTML or te

Search the persisted Wirecutter recommendation database by product type, pick name, review title, category, and model tokens. Search is lexical today; pass concise, spelling-corrected product/category nouns instead of vague natural-language intent.

Answers questions (single-turn question answering) from a small local LLM (qwen3.5 2B). Best for general knowledge, definitions and simple reasoning. POST JSON {"input": string}, up to 3000 characters; returns {answer, model, elapsed_ms}. Answers are brief (2-3 sentences). Typically 5-15 s. Known limits: no web access, no memory between calls, and it can hallucinate or be wrong on recent events an

Eight quarters of a US public company's reported fundamentals, straight from SEC XBRL. Answers "Apple revenue last 8 quarters", "is NVDA net income growing", "MSFT EPS trend". Revenue, net income, diluted EPS per quarter with computed YoY changes and the exact XBRL tag used — every number verbatim from data.sec.gov companyfacts, nothing modeled. Accepts ticker, company name, or CIK. No LLM, no key

Returns clean, normalized datasets of AI model benchmarks, pricing, leaderboard scores, and Ethereum EIPs, parsed from sources that are costly to scrape.

Converting Apple Immersive (AIVU) for Pico headsets with correct geometry. A working AIVU-to-Pico pipeline that renders through Apple's own calibration mesh (not a generic fisheye model), outputs 360 equirect stereo SBS for ArborXR, with exact FFmpeg commands, API calls, the traps that scramble or squish the image, and honest quality limits. Tested on Pico 4 Enterprise and Pico 4 Ultra.

LLM chat completion with no API key or signup (fast open model, meta-llama/llama-3.3-70b-instruct). OpenAI-compatible: set your SDK base URL to https://keyless.run/fast/v1. Flat $0.002 per call, up to 1500 output tokens. Charged only on success.

Jev hires a model for your task and grades the result. Rejected work is refunded.

Recent transaction history for an address, normalized across Ethereum, Base, Arbitrum, Optimism, Polygon, Bitcoin, and Solana -- direction, counterparty, and value where the chain's data model makes that derivable.

US safety recalls for a VIN, deterministic. Answers "does this car have open recalls", "is this used car safe to buy", "what NHTSA campaigns cover this vehicle". Decodes the VIN via NHTSA vPIC (year, make, model, body, engine) then lists every NHTSA recall campaign for that vehicle: component, summary, remedy, park-outside and do-not-drive flags. No LLM, no key, public domain.

Returns hourly surface concentrations from the CAMS European air-quality ensemble for a point in the cached DACH and Central Europe area. Available pollutants are CO, NO2, NO, O3, PM2.5, PM10 and SO2. Values come from the nearest 0.1 degree model cell and are labelled as a regional forecast rather than a street-level sensor reading. The local cutout is refreshed independently of paid calls, so Cop

AI crawler visibility audit: is your website visible to AI crawlers (GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot) so ChatGPT, Claude and Perplexity can read and cite it? Fetched like an AI crawler, without JavaScript: robots.txt, llms.txt, sitemap, soft 404, JSON-LD, and which crawlers are blocked at the edge. $0.01 per check, one payment per check: every rule, a model verdict, a fix with evid

Format, dimensions, megapixels, bytes, channels, transparency, color space, bit depth, DPI, orientation, animation and pages, ICC profile, and EXIF camera details (make, model, date taken, exposure, aperture, ISO, focal length, GPS position) of an image at a URL.

Cheapest cloud GPU-hour for a GPU model (H100, H200, B200, A100, MI300X, L40S, L4, A10G, A10, V100, T4) across AWS and Azure regions, on-demand or spot, plus per-provider summary and 30-day price history.

AI Voice - Pharma - Rheumatology: Psoriatic Arthritis & Ankylosing Spondylitis - Argentina 2026 (mini) (Voz IA - Farma - Reumatologia: artritis psoriasica y espondilitis anquilosante - Argentina 2026 (mini)). Full study in one call: question-universe design, 114 raw AI answers (4 models), 7 analysis tables with Lightning metrics, 20 findings with evidence. Spanish-language data from Argentina, run

GPT-5.6 Luna — fast, low-cost frontier model (Artificial Analysis Intelligence Index 51.2, GPQA Diamond 91.1%). Best for high-volume chat, classification, extraction, and lightweight agentic loops where latency and cost matter. Limits: 8K input / 2K output tokens per call. OpenAI-compatible request and response format.

Structured evaluation with TypeSafe's Jev (System One) model. Send any state plus typed questions — noul (true/false), choice (pick one), or score (ordinal scale) — and get back calibrated probabilities and confidence for each. Built for routing, classification, triage and policy checks where a fast, predictable, machine-readable answer beats prose.

Structured evaluation with TypeSafe's Jev (System One) model. Send any state plus typed questions — noul (true/false), choice (pick one), or score (ordinal scale) — and get back calibrated probabilities and confidence for each. Built for routing, classification, triage and policy checks where a fast, predictable, machine-readable answer beats prose.

Generate AI video with Base USDC. Price depends on model, duration, and resolution. Returns a free status URL to poll.

Research cards from KARP, SoulDriver's multi-model research engine: GPT, Gemini, Grok and Claude debate a topic over many rounds and publish a conclusion. Each card gives the conclusion, key insight, warnings, dated figures, consensus status, open questions, research date and staleness. See what is covered first with the free POST /v1/coverage; verify time-sensitive points with POST /v1/check. Cat

URL-to-Markdown web extraction and context compression for AI agents. Fetch a public web page and return clean, task-relevant, citation-verifiable LLM context with exact excerpts, locators, hashes, and estimated tokens saved. Use before an agent summarizes, answers questions from, or cites a long or noisy URL. Deterministic and model-free. Public HTML only; no PDFs, logins, or JavaScript-only page

Remove the background from a photo. Send an image URL, get a transparent PNG cutout of the subject (BiRefNet model).

LLM cost scout: send your monthly OpenAI API spend in USD and get back the flat ChatGPT/Codex subscription tier that absorbs it through the ProxyLLM gateway, the exact new monthly total and savings, the live model catalog, and machine-readable signup steps an agent can complete autonomously (email OTP, no browser).

Give this API any text or binary data and it returns its fingerprint under whichever hash algorithms you choose - SHA-256, SHA-512, MD5, BLAKE2, CRC32 and more. It also signs data with a secret key (HMAC) and safely verifies whether an incoming signature matches, using a timing-safe comparison. Language models cannot hash: they emit a string of the right length and character set that is simply wro

TEXT-IN ONLY sentiment trajectory: per-item score, inflection points, drift magnitude + model version. Not a trading signal. No scrape / no X. OpenRouter free-first.

Check NHTSA safety recall campaigns for a vehicle by make, model, and year — or by VIN (decoded live via NHTSA vPIC, then matched). Returns campaign numbers, components, defect/consequence/remedy summaries as published, report dates, units affected, and do-not-drive / park-outside advisories. Recall STATUS only — not a VIN decoder and not a per-VIN remedy-completion check. US federal public data (

Fetches live offers from a public bridge aggregator and reports for every option what arrives, what it costs in fees and gas, and how long it usually takes. The security model is given equal weight to the price: bridges are the most expensive failures this industry has, so trust_assumptions spells out who a user has to trust for that route to hold. No recommendation is forced - three best values b

Can I sell this Base token? Honeypot detector and rug check: simulates a real buy and sell (Uniswap V2/V3/V4 incl. Zora, Clanker, Doppler) and returns is_honeypot, buy/sell tax, exit depth, mint/blacklist/pause/fee powers, owner, proxy, deployer history, holders, token age, a verdict + score + model odds. GoPlus-compatible fields included.

Verify output another AI agent hands back, by code alone, for $0.005: no model call, answers in about a second. Recomputes the deliverable's written sums, totals, percentages, dates, weekdays, time spans, unit conversions, and list counts; checks lengths the task asks for; flags placeholders and prompt injection; measures words, items, and JSON shape. Signed receipt, tier lite. For every requireme