From 0d608c1d7c4c8eb266669084b7a6f2f51064b073 Mon Sep 17 00:00:00 2001 From: 1bcMax Date: Sun, 15 Nov 2026 18:30:55 +0000 Subject: [PATCH] docs(readme): resync model tables with live catalog; market data, Surf, Live Search, prices - Add catalog models missing from the tables (claude-fable-5.2, gemini-4-carbon, kimi-k3.1-pro, deepseek-v4.1-flash, gpt-6.1-sol, qwen3.8-max, mistral-large-4, hy4-preview, laguna-s-2.1, free minimax-m3.1-flash) and the new image models; drop retired gpt-4.1-nano and dall-e-3. - Free tier: list the pin-worthy free models, warn not to pin the four ids the gateway now serves with another model, fix the free routing-profile row. - PriceClient: free daily closes (Chainlink/Backpack), not Pyth or realtime; equity price/history answer 501, raised as APIError(status_code=501). - Surf: /v1/surf/* is retired (410), SurfClient deprecated. - Live Search example used openai/gpt-5.2 (400 SEARCH_UNSUPPORTED); use xai/grok-4.7. - Drop per-call prices the live 402 quotes no longer match; fix Modal pricing. --- CHANGELOG.md | 30 ++++++++++ README.md | 161 +++++++++++++++++++++++++++++---------------------- 2 files changed, 123 insertions(+), 68 deletions(-) diff --git a/CHANGELOG.md b/CHANGELOG.md index 2086aca..5456040 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -4,6 +4,36 @@ All notable changes to blockrun-llm will be documented in this file. ## Unreleased +### Docs +- **README model tables re-synced with the live catalog** (`GET /v1/models`, + 2026-11-15). Added `anthropic/claude-fable-5.2`, `google/gemini-4-carbon`, + `moonshot/kimi-k3.1-pro` (replaces `kimi-k3`, which is hidden but still + callable), `deepseek/deepseek-v4.1-flash`, `openai/gpt-6.1-sol`, + `qwen/qwen3.8-max`, `mistral/mistral-large-4`, `tencent/hy4-preview`, + `poolside/laguna-s-2.1` and the free `nvidia/minimax-m3.1-flash`. Removed the + retired `openai/gpt-4.1-nano` and `openai/dall-e-3`; added the image models + the catalog lists (`gpt-image-2.5-flare` / `-sunburst`, `nano-banana-2`, + `grok-imagine-image-2.0`, `seedream-5-pro`). +- **Free tier:** the README names the free models worth pinning and says not + to pin `nvidia/llama-3.2-11b-vision`, `nvidia/nemotron-3-super-120b`, + `cohere/north-mini-code` or `nvidia/muse-glimmer-30b`, which the gateway now + serves with a different model. The `free` routing-profile row lists the + models `routing_profile="free"` actually routes across. +- **Market data:** `PriceClient` docs no longer say Pyth or realtime. Crypto, + FX and commodity price, history and lists are free daily closes (Chainlink / + Backpack, `realtime: false`). Equity price and history answer 501 on both + gateways, which `PriceClient` raises as `APIError` with `status_code == 501`; + stock symbol lists still work. +- **Surf:** the README section now says `/v1/surf/*` is retired (410) and + `SurfClient` is deprecated, and points to the replacements. +- **Live Search:** the `search=True` example used `openai/gpt-5.2`, which the + gateway refuses with 400 `SEARCH_UNSUPPORTED`; it now uses `xai/grok-4.7`. +- Dropped per-call prices the live 402 quotes no longer match (Live Search, + `SearchClient`, Predexon tiers, sound effects, the paid DeepSeek V4 Pro rate, + the FAQ's paid-model floor) and the unverifiable testnet prices; corrected + the Modal sandbox pricing to match the gateway. The FAQ no longer says the + SDK takes no API keys. + ## 1.22.3 — 2026-11-14 ### Fixed diff --git a/README.md b/README.md index 1f4e036..c7eb781 100644 --- a/README.md +++ b/README.md @@ -1,6 +1,6 @@ # BlockRun LLM SDK (Python) -> **blockrun-llm** is a Python SDK for accessing 91 large language models (GPT-6, GPT-5.x, Claude 5.x, Gemini 3.x, DeepSeek, Grok 4.x, GLM, MiniMax, Moonshot and more) plus image / video / music generation, Grok Live Search, prediction-market data (Predexon), Exa neural web search, and Pyth-backed market data. Every call is paid per request — no subscription, no seats, no minimum. Built for AI agents that need to operate autonomously. +> **blockrun-llm** is a Python SDK for accessing 91 large language models (GPT-6, GPT-5.x, Claude 5.x, Gemini 4 / 3.x, DeepSeek, Grok 4.x, GLM, MiniMax, Moonshot and more) plus image / video / music generation, Grok Live Search, prediction-market data (Predexon), Exa neural web search, and free crypto / FX / commodity price data (daily closes). Paid calls are billed per request — no subscription, no seats, no minimum. Built for AI agents that need to operate autonomously. > > **Two ways to pay, same SDK, same catalogue.** Sign up at > **[user.blockrun.ai](https://user.blockrun.ai)** for an API key and prepaid @@ -8,7 +8,7 @@ > request settle itself over x402 — on **Solana or Base**. Every client takes > either credential in the same first argument. > -> 🆓 **Includes 9 free models** — Nemotron 3.5 Lightning (1M context), Nemotron 3 Ultra 550B, Nemotron 3 Nano Omni, Llama 3.2 11B Vision, Cohere North Mini Code and Poolside Laguna XS 2.1. Zero USDC, no rate-limit gimmicks. Use `routing_profile="free"` or call any of them directly. +> 🆓 **Includes 9 free models** — Nemotron 3.5 Lightning (1M context), Nemotron 3 Ultra 550B, MiniMax M3.1 Flash (1M context), Nemotron 3 Nano Omni, Nemotron 3.5 Bolt (vision) and Poolside Laguna XS 2.1. Zero USDC, no rate-limit gimmicks. Use `routing_profile="free"` or pin one of the models listed under [Try It Free](#try-it-free-no-balance-required). [![PyPI](https://img.shields.io/pypi/v/blockrun-llm.svg)](https://pypi.org/project/blockrun-llm/) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) @@ -85,13 +85,16 @@ print(result.response) # '4' |----------|---------|----------| | `nvidia/nemotron-3.5-lightning` | 1M | The free default — thinking-mode reasoning, longest free context | | `nvidia/nemotron-3-ultra-550b` | 1M | Largest free model (550B / 55B active MoE) — strongest, but slower | +| `nvidia/minimax-m3.1-flash` | 1M | MiniMax M3.1 Flash — reasons before it answers, calls tools, usually answers in a few seconds | | `nvidia/nemotron-3-nano-omni-30b-a3b-reasoning` | 256K | Fast general chat + reasoning (30B-A3B) | | `nvidia/nemotron-3.5-bolt` | 131K | Free vision — verified on real image probes on Base and Solana | | `poolside/laguna-xs-2.1` | 131K | Coding, ~161 tok/s | -> For free image input use `nemotron-3.5-bolt`. `nemotron-3-nano-omni` is catalogued as vision-capable but failed earlier image probes on Solana, so the router never sends it images. `nvidia/llama-3.2-11b-vision` and `cohere/north-mini-code` are no longer served as themselves: the gateway answers them with another model, so do not pin them. +> For free image input use `nemotron-3.5-bolt`. `nemotron-3-nano-omni` is catalogued as vision-capable but failed earlier image probes on Solana, so the router never sends it images. -> Need V4-Pro-class reasoning? Use the paid `deepseek/deepseek-v4-pro` ($0.435/$0.87 — the 75% launch promo became the permanent list price after 2026-05-31) — the free `nvidia/deepseek-v4-pro` is retired, and the gateway redirects calls to it to a free NVIDIA Nemotron model. +> **Do not pin these free ids:** `nvidia/llama-3.2-11b-vision`, `nvidia/nemotron-3-super-120b`, `cohere/north-mini-code` and `nvidia/muse-glimmer-30b`. They still answer 200, but since 2026-11 the gateway serves them with a different model, so a pinned call does not get the model it names. + +> Need V4-Pro-class reasoning? Use the paid `deepseek/deepseek-v4-pro`. The free `nvidia/deepseek-v4-pro` is retired and no longer in the catalog; calls to it are redirected to a free NVIDIA Nemotron model. > **Retired**: NVIDIA has EOL'd most of its early free lineup — the free DeepSeek family (last: `nvidia/deepseek-v4-flash`, 2026-08-12), `step-3.7-flash`, `mistral-nemotron`, `nemotron-nano-9b-v2` / `-12b-v2-vl` (2026-08-30), `gpt-oss-120b/20b` (2026-09-03), `nemotron-3-nano-30b` (2026-09-08), `llama-4-maverick`, the qwen3 SKUs, free Mistral small/large, and more. The gateway auto-redirects pinned callers to a live model, so old model IDs still return 200 — from a different model. @@ -263,7 +266,7 @@ print(decision.reasoning) # human-readable explanation of the pick | Profile | Description | Best For | |---------|-------------|----------| -| `free` | Free tier — smart-routes across the $0 models (Nemotron 3 Nano Omni, Nemotron 3.5 Lightning, Llama 3.2 11B Vision, Cohere North Mini Code, Poolside Laguna XS 2.1). Nemotron 3 Ultra 550B is free too, by direct call. | Zero-cost testing, dev, prod | +| `free` | Free tier — smart-routes across the $0 models (Nemotron 3 Nano Omni, Nemotron 3.5 Lightning, Nemotron 3.5 Bolt, Nemotron 3 Ultra 550B, Poolside Laguna XS 2.1). MiniMax M3.1 Flash is free too, by direct call. | Zero-cost testing, dev, prod | | `eco` | Cheapest capable model per tier | Cost-sensitive production | | `auto` | Best balance of cost/quality (default) | General use | | `premium` | Top-tier models (Anthropic, OpenAI, Moonshot) | Quality-critical tasks | @@ -525,12 +528,13 @@ and `client.list_image_models()` return exactly what the gateway is charging. ### OpenAI GPT-6 Family The GPT-6 generation: Astra is the flagship for long-horizon agentic work and -computer use, Sol the cost-efficient tier for complex coding, Luna the fast -low-cost tier. +computer use, Sol the cost-efficient tier for complex coding (`gpt-6.1-sol` is the +refreshed Sol), Luna the fast low-cost tier. | Model | Context | |---|---| | `openai/gpt-6-astra` | 1.05M | +| `openai/gpt-6.1-sol` | 1.05M | | `openai/gpt-6-sol` | 1.05M | | `openai/gpt-6-luna` | 1.05M | @@ -572,7 +576,6 @@ longer at the same token price. |---|---| | `openai/gpt-4.1` | 128K | | `openai/gpt-4.1-mini` | 128K | -| `openai/gpt-4.1-nano` | 128K | | `openai/gpt-4o` | 128K | | `openai/gpt-4o-mini` | 128K | @@ -589,7 +592,8 @@ longer at the same token price. | Model | Context | Notes | |---|---|---| -| `anthropic/claude-fable-5.1` | 1M | Most capable — successor to Fable 5 at the same tier and price | +| `anthropic/claude-fable-5.2` | 1M | Most capable — successor to Fable 5.1 at a lower price | +| `anthropic/claude-fable-5.1` | 1M | Successor to Fable 5 at the same tier and price | | `anthropic/claude-fable-5` | 1M | Mythos-class flagship above Opus — always-on thinking, 128K output | | `anthropic/claude-opus-5.5` | 1M | Newest Opus — Opus-class reasoning at a lower price than Opus 5 | | `anthropic/claude-opus-5` | 1M | Flagship — the baseline the routing savings claim is measured against | @@ -606,6 +610,7 @@ longer at the same token price. | Model | Context | |---|---| +| `google/gemini-4-carbon` | 1M | | `google/gemini-4-argon` | 1M | | `google/gemini-3.1-pro` | 1M | | `google/gemini-3.8-flash` | 1M | @@ -622,11 +627,13 @@ longer at the same token price. DeepSeek upstream serves the legacy `deepseek-chat` / `deepseek-reasoner` aliases as V4 Flash non-thinking / thinking modes. V4 Pro is the flagship -paid SKU; the vision SKU is an experimental preview. +paid SKU; V4.1 Flash is the newer Flash in non-thinking mode; the vision SKU +is an experimental preview. | Model | Context | Notes | |---|---|---| | `deepseek/deepseek-v4-pro` | 1M | V4 flagship — strongest open-weight reasoner | +| `deepseek/deepseek-v4.1-flash` | 1M | V4.1 Flash, non-thinking mode | | `deepseek/deepseek-v4-flash-vision-exp` | 1M | Experimental vision preview | | `deepseek/deepseek-chat` | 1M | V4 Flash non-thinking | | `deepseek/deepseek-reasoner` | 1M | V4 Flash thinking (same upstream, thinking on by default) | @@ -650,7 +657,7 @@ only ranks what `/v1/models` lists. | Model | Context | Notes | |---|---|---| -| `moonshot/kimi-k3` | 1M | Replaces the retired `kimi-k2.5` / `k2.6` SKUs | +| `moonshot/kimi-k3.1-pro` | 1M | Moonshot flagship — successor to Kimi K3; image + text input, returns `reasoning_content`. `kimi-k3` is hidden from the catalog but still callable by id | | `minimax/minimax-m3` | 1M | | | `minimax/minimax-m2.7` | 200K | | | `zai/glm-5.3` | 1M | | @@ -659,6 +666,7 @@ only ranks what `/v1/models` lists. | `zai/glm-5.1` | 200K | | | `zai/glm-5` | 200K | | | `zai/glm-5-turbo` | 200K | | +| `qwen/qwen3.8-max` | 1M | Current Qwen flagship | | `qwen/qwen3.7-max` | 1M | | | `qwen/qwen3.7-plus` | 1M | | | `qwen/qwen3.8-flash` | 1M | | @@ -671,6 +679,14 @@ only ranks what `/v1/models` lists. | `xiaomi/mimo-v2.5` | 1M | | `xiaomi/mimo-v2.5-pro` | 1M | +### Mistral, Tencent, Poolside + +| Model | Context | Notes | +|---|---|---| +| `mistral/mistral-large-4` | 512K | Mistral flagship — reasoning, coding, image input | +| `tencent/hy4-preview` | 1M | Hy4 — 770B open-weight MoE (Apache 2.0) | +| `poolside/laguna-s-2.1` | 1M | 118B open-weight coding model | + ### Free Tier The 9 free models are listed under @@ -712,13 +728,17 @@ automatically. | Model | Notes | | ------- | --- | -| `openai/dall-e-3` | | | `openai/gpt-image-1` | | | `openai/gpt-image-2` | reasoning-driven, multilingual text rendering, character consistency | +| `openai/gpt-image-2.5-flare` | | +| `openai/gpt-image-2.5-sunburst` | | | `google/nano-banana` | | +| `google/nano-banana-2` | | | `google/nano-banana-pro` | | | `xai/grok-imagine-image` | | +| `xai/grok-imagine-image-2.0` | | | `xai/grok-imagine-image-pro` | | +| `bytedance/seedream-5-pro` | | | `zai/cogview-4` | | Image editing (`client.edit` / `client.image_edit`) hits the `/v1/images/image2image` endpoint and supports `openai/gpt-image-1`, `openai/gpt-image-2`, `google/nano-banana`, and `google/nano-banana-pro`. Pass a list of source images to fuse multiple inputs (openai/* up to 4, google/* up to 3). @@ -840,7 +860,7 @@ result = client.generate( speed=1.1, ) -# Sound effects (flat $0.05/generation) +# Sound effects (flat price per generation) result = client.sound_effect("rain on a tin roof, distant thunder") # List voices (free, rate-limited) @@ -1015,39 +1035,26 @@ client.renew_number(bought["phone_number"]) # $5.00, +30 days client.release_number(bought["phone_number"]) # free ``` -## Surf — Crypto Intelligence (`SurfClient`) - -`SurfClient` wraps `/v1/surf/*` — the asksurf.ai partner gateway, ~83 crypto -endpoints across exchanges, on-chain SQL, prediction markets (Polymarket + -Kalshi), wallets, social analytics, and project intelligence. Tiered pricing: -$0.001 / $0.005 / $0.020 per call (tier 1 / 2 / 3). +## Surf (`SurfClient`) — retired -```python -from blockrun_llm import SurfClient - -client = SurfClient() - -# Discovery -print(SurfClient.endpoints()) # full catalog -print(client.price("market/ranking")) # 0.001 -print(client.endpoint_info("onchain/sql")) # {'method': 'POST', 'tier': 3, ...} +**Deprecated.** The gateway retired `/v1/surf/*` on 2026-09-06: every Surf +endpoint answers `410 endpoint_retired` on both blockrun.ai and sol.blockrun.ai, +before any payment is asked for. `SurfClient` is still importable, but +constructing it emits a `DeprecationWarning`, and every call that reaches the +gateway fails with that 410 as an `APIError`. It will be removed in a future release. -# GET — pass query params (validated against the catalog) -btc_price = client.get("exchange/price", {"pair": "BTC/USDT"}) -holders = client.get("token/holders", {"address": "0x...", "chain": "ethereum"}) +Use these instead: -# POST — JSON body -rows = client.post("onchain/sql", {"query": "SELECT count() FROM ethereum.blocks"}) - -# Generic helper — auto-routes GET vs POST from the catalog -result = client.call("token/holders", params={"address": "0x...", "chain": "ethereum"}) -``` +- Prices: [`PriceClient`](#market-data-priceclient) (crypto, FX, commodities). +- Prediction markets: [`client.pm_*`](#prediction-markets-powered-by-predexon-v2). +- Protocol TVL and yields: [`client.defi_*`](#defi-data-powered-by-defillama). +- DEX quotes and swaps: [`client.dex_*`](#dex-swaps-powered-by-0x). ## Standalone Search (`SearchClient`) `SearchClient` wraps `POST /v1/search` — standalone Grok Live Search with -automatic x402 payment. Pricing: `$0.025/source + margin` -(10 sources ≈ `$0.26`). +automatic x402 payment. The exact price is in each call's 402 quote before +you pay. ```python from blockrun_llm import SearchClient @@ -1065,34 +1072,47 @@ for url in result.citations or []: ## Market Data (`PriceClient`) -Pyth-backed realtime quotes and OHLC history across crypto, FX, commodities -and 12 global equity markets. Crypto / FX / commodity are **fully free** -across price, history and list; stocks (`stocks/{market}` and the `usstock` -legacy alias) charge `$0.001` per price or history call. Pass -`require_wallet=False` when you only need free endpoints. +Crypto, FX and commodity prices, OHLC history and symbol lists — all **free**, +no payment and no wallet needed. Prices are **daily closes, not realtime**: +responses carry `realtime: false` (`delayed: true` on sol.blockrun.ai) and name +their `source` (Chainlink or Backpack). Pass `require_wallet=False` so the +client does not insist on a wallet it never uses. ```python +import time from blockrun_llm import PriceClient -# Free usage — no wallet p = PriceClient(require_wallet=False) + btc = p.price("crypto", "BTC-USD") +print(btc.price, btc.source, btc.realtime) # e.g. 64712.9 backpack False eur = p.price("fx", "EUR-USD") -symbols = p.list_symbols("crypto", q="sol", limit=20) +gold = p.price("commodity", "XAU-USD") -# Paid — requires a wallet -p2 = PriceClient() -aapl = p2.price("stocks", "AAPL", market="us") -bars = p2.history( - "stocks", "AAPL", - market="us", - resolution="D", - from_ts=1_700_000_000, - to_ts=1_710_000_000, -) +now = int(time.time()) +bars = p.history("crypto", "BTC-USD", resolution="D", + from_ts=now - 30 * 86400, to_ts=now) +symbols = p.list_symbols("crypto", q="sol", limit=20) ``` -Supported stock markets: `us, hk, jp, kr, gb, de, fr, nl, ie, lu, cn, ca`. +**Equities are not served.** `stocks/{market}` and the `usstock` alias still +list symbols (`p.list_symbols("stocks", market="us")` answers normally; markets +`us, hk, jp, kr, gb, de, fr, nl, ie, lu, cn, ca`), but price and history answer +HTTP **501** ("We do not currently serve equity +prices") on both blockrun.ai and sol.blockrun.ai, without a payment request. +`PriceClient` raises that as an `APIError` with `status_code == 501` and the +gateway's error title as the message (e.g. +`API error: 501: US Stock price is not available`): + +```python +from blockrun_llm import APIError + +try: + p.price("stocks", "AAPL", market="us") +except APIError as e: + if e.status_code == 501: + ... # equity prices are not available from the gateway +``` ## Multi-chain RPC (`RpcClient`) @@ -1183,7 +1203,10 @@ client.dex_gasless_chains() # supported gasless chains ## Cloud Compute (Powered by Modal) Pay-per-call sandboxed compute — create a sandbox, run commands, tear it -down. $0.01/create (CPU; $0.05 with GPU), $0.001 per exec/status/terminate. +down. A small sandbox (up to 300 s, 1 CPU, 1024 MB) is a flat $0.01 to create +on CPU, $0.05-$0.40 with a GPU; a longer-lived sandbox is billed hourly for +its requested lifetime, upfront. Exec / status / terminate are $0.001 each. +The 402 quote is the exact price. ```python sb = client.modal_sandbox_create(image="python:3.11") @@ -1207,7 +1230,7 @@ print(link["url"]) # https://pay.coinbase.com/... — open to buy USDC on Base ## Prediction Markets (Powered by Predexon v2) -Access real-time prediction market data from Polymarket, Kalshi, Limitless, sports, and Binance Futures via [Predexon](https://predexon.com). No API keys needed — pay-per-request via x402. Tier 1 endpoints are $0.001/call, Tier 2 (wallet identity / clustering) are $0.005/call. +Access real-time prediction market data from Polymarket, Kalshi, Limitless, sports, and Binance Futures via [Predexon](https://predexon.com). No Predexon account needed — pay per request with an API key or x402. Each call's exact price is in its 402 quote before you pay; live rates are at [blockrun.ai/pricing](https://blockrun.ai/pricing). Each method below is available on `LLMClient` (Base), `AsyncLLMClient`, and `SolanaLLMClient`. @@ -1261,7 +1284,7 @@ cluster = client.pm_wallet_cluster("0xABC123...") ### Generic passthrough For endpoints without a typed helper, drop down to `pm()` (GET) or `pm_query()` -(POST). Same pricing tiers, same return shape: +(POST). Same pricing, same return shape: ```python candles = client.pm("polymarket/candlesticks/0x1234abcd...") # OHLCV @@ -1423,17 +1446,19 @@ from blockrun_llm import LLMClient client = LLMClient() -# Simple: Enable live search with search=True (default 10 sources, ~$0.26) +# Live Search runs on xai/* models only; any other model answers +# 400 SEARCH_UNSUPPORTED. +# Simple: enable live search with search=True response = client.chat( - "openai/gpt-5.2", + "xai/grok-4.7", "What are the latest posts from @blockrunai?", search=True ) print(response) -# Custom: Limit sources to reduce cost (5 sources, ~$0.13) +# Custom: full search_parameters (the 402 quote states the price before you pay) response = client.chat( - "openai/gpt-5.2", + "xai/grok-4.7", "What's trending on X?", search_parameters={"mode": "on", "max_search_results": 5} ) @@ -1537,8 +1562,8 @@ print(f"Testnet USDC: ${balance:.4f}") ### Available Testnet Models -- `openai/gpt-oss-20b` - $0.001/request (flat price) -- `openai/gpt-oss-120b` - $0.002/request (flat price) +- `openai/gpt-oss-20b` (flat price per request) +- `openai/gpt-oss-120b` (flat price per request) ### Manual Testnet Configuration @@ -1984,7 +2009,7 @@ The `AnthropicClient` wraps `anthropic.Anthropic` with a custom httpx transport ## Frequently Asked Questions ### What is blockrun-llm? -blockrun-llm is a Python SDK that provides pay-per-request access to 91 large language models from OpenAI, Anthropic, Google, DeepSeek, NVIDIA, ZAI, and more. It uses the x402 protocol for automatic USDC micropayments — no API keys, no subscriptions, no vendor lock-in. +blockrun-llm is a Python SDK that provides pay-per-request access to 91 large language models from OpenAI, Anthropic, Google, DeepSeek, NVIDIA, ZAI, and more. Pay with a BlockRun API key (prepaid credit from [user.blockrun.ai](https://user.blockrun.ai)) or with automatic x402 USDC micropayments from your own wallet — no subscriptions, no vendor lock-in. ### How does payment work? When you make an API call, the SDK automatically handles x402 payment. It signs a USDC transaction locally using your wallet private key (which never leaves your machine), and includes the payment proof in the request header. Settlement is non-custodial and instant on Base or Solana. @@ -1993,7 +2018,7 @@ When you make an API call, the SDK automatically handles x402 payment. It signs Router Core is BlockRun's built-in routing engine — shared with the TypeScript SDK and the gateway, so the same request routes the same way everywhere. It scores your request across 15 dimensions, drops every model that can't actually handle it (context, output length, tools, vision), then picks the cheapest capable one and keeps the rest as a fallback chain. Routing happens locally in under 1ms and makes no extra model call. It can save up to 84% on LLM costs compared to using premium models for every request. ### How much does it cost? -Pay only for what you use. Prices start at **FREE** (9 free models). Paid models start at $0.10/M tokens. There are no minimums, subscriptions, or monthly fees. $5 in USDC gets you thousands of requests. +Pay only for what you use. Prices start at **FREE** (9 free models). Live per-model prices are at [blockrun.ai/pricing](https://blockrun.ai/pricing). There are no minimums, subscriptions, or monthly fees. $5 in USDC gets you thousands of requests. ### Can I use it with Solana? Yes. Install with `pip install blockrun-llm[solana]` and use `SolanaLLMClient` instead of `LLMClient`. Same API, different payment chain.