Name: BrowseWiz
Brand: BrowseWiz
Availability: InStock

https://openrouter.ai/api/v1/chat/completions

. OpenRouter accepts the OpenAI Chat Completions format, so the official OpenAI SDKs work once you set the base URL to that address, send your OpenRouter key as a Bearer token and name a model such as 

 and the same request goes to OpenAI or Mistral. Vercel AI Gateway, Portkey and the self-hosted LiteLLM Proxy also put Anthropic, OpenAI and Mistral behind one OpenAI-compatible endpoint; they're compared below.

Every URL, model ID and price on this page was checked against the vendors' own documentation and APIs in September 2026.

OpenRouter API base URL and chat completions endpoint

The optional headers only attribute your app: 

 is the name shown in OpenRouter's rankings, and 

 assigns marketplace categories. Requests work without them.

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -d '{
    "model": "anthropic/claude-sonnet-5",
    "messages": [
      {"role": "user", "content": "Explain what an OpenAI-compatible API is in one sentence."}
    ]
  }'

Python example with the official openai SDK

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://openrouter.ai/api/v1",
    api_key=os.environ["OPENROUTER_API_KEY"],
)

completion = client.chat.completions.create(
    model="anthropic/claude-sonnet-5",
    messages=[
        {"role": "user", "content": "Explain what an OpenAI-compatible API is in one sentence."}
    ],
    extra_headers={
        "HTTP-Referer": "https://example.com",  # optional: your site URL
        "X-OpenRouter-Title": "My App",          # optional: your app name
    },
)

print(completion.model)
print(completion.choices[0].message.content)

 itself, so if you paste the full endpoint as 

Model IDs and switching between Claude, GPT and Mistral

). The endpoint, key and payload stay the same, so switching vendors is a loop over strings, reusing the 

for model in ["openai/gpt-6-sol", "anthropic/claude-sonnet-5", "mistralai/mistral-medium-3-5"]:
    r = client.chat.completions.create(
        model=model,
        messages=[{"role": "user", "content": "Name one risk of vendor lock-in."}],
    )
    print(f"{r.model}: {r.choices[0].message.content}")

Two OpenRouter features help once you switch often. Aliases of the form 

, resolve to the newest model in that family, and the response's 

 field names the model that actually ran. A 

 array lists fallbacks that OpenRouter tries in order when the first model fails on rate limits, downtime, context length or a moderation flag. The OpenAI SDK has no argument for it, so pass it as 

; you are billed for the model that answered.

The full catalog, with the parameters each model supports, is at 

, which needs no key. If you're still choosing what to route to, our 

On pricing, OpenRouter passes provider prices through with no markup on inference and takes a 5.5% fee (minimum $0.80) when you buy credits by card, or 5% for crypto.

Services that let you switch between Anthropic, OpenAI and Mistral with one API

As of September 2026, OpenRouter, Vercel AI Gateway, Portkey and the self-hosted LiteLLM Proxy each put Anthropic, OpenAI and Mistral models behind one OpenAI-compatible endpoint. Cloudflare AI Gateway's current endpoint does the same for Anthropic, OpenAI and Google. For Mistral it offers only open-weight models on Workers AI; Mistral's own API needs the older 

 endpoint, which Cloudflare now deprecates for single-model calls. In every case you switch vendors by changing the 

 string. The vendors also run their own OpenAI-compatible endpoints, which help when you want one client library and don't mind one key per vendor.

If you don't want an account with every vendor, OpenRouter, Vercel AI Gateway and Cloudflare's Unified Billing sell tokens at provider prices from one prepaid balance. When we checked (September 2026), OpenRouter's models API listed 356 distinct models, not counting aliases and 

 variants, against 265 language models on Vercel's, so it is the pick when you want breadth. Vercel and Cloudflare make more sense when your app already runs on them.

If you already hold vendor keys, Portkey and LiteLLM sit in front of them. Portkey adds retries, fallbacks, load balancing and guardrails, and its free tier covers 10,000 recorded logs a month. LiteLLM tracks spend and sets budgets per virtual key, and it is the option when no third-party gateway may see your traffic.

Model IDs don't carry over between gateways. OpenRouter writes 

 you define. The Anthropic and Gemini compatibility layers are the weakest options for production: Anthropic describes its layer as a way to test and compare models, not a long-term solution, and Google still labels its layer beta.

A LiteLLM config that exposes Claude, GPT and Mistral under short names:

model_list:
  - model_name: claude
    litellm_params:
      model: anthropic/claude-sonnet-5
      api_key: os.environ/ANTHROPIC_API_KEY
  - model_name: gpt
    litellm_params:
      model: openai/gpt-6-sol
      api_key: os.environ/OPENAI_API_KEY
  - model_name: mistral
    litellm_params:
      model: mistral/mistral-large-latest
      api_key: os.environ/MISTRAL_API_KEY

docker run -v $(pwd)/litellm_config.yaml:/app/config.yaml \
  -e ANTHROPIC_API_KEY -e OPENAI_API_KEY -e MISTRAL_API_KEY \
  -e LITELLM_MASTER_KEY=sk-change-me \
  -p 4000:4000 docker.litellm.ai/berriai/litellm:latest --config /app/config.yaml

. Without a database, LiteLLM does not enforce budgets.

What "OpenAI-compatible" means, and where it breaks

), authenticates with a Bearer token and returns the same 

 shape, or server-sent events when streaming. Plain chat requests work across every endpoint in the table once you change the base URL, key and model name. The differences sit at the edges:

Write against the common subset (messages, streaming, tools without 

) and keep vendor-specific parameters in 

Using any OpenAI-compatible model in BrowseWiz

BrowseWiz is an AI side-panel assistant for Chrome and Edge, and its Custom Models setting accepts any model that exposes an OpenAI API-compatible chat completions endpoint. It asks for the same three values as the SDK:

Each entry is one model. To use Claude, GPT and Mistral through OpenRouter, add three entries with the same base URL and key and different model names. Direct vendor endpoints work the same way:

If you only need OpenAI models, the Bring Your Own Key field on the same page takes an OpenAI key directly. Without any key, the built-in BrowseWiz Chat model works on the free plan with limited usage. The settings page also takes custom API and webhook tools; our 

Authorization: Bearer <OPENROUTER_API_KEY>

Yes. OpenRouter calls itself a drop-in replacement for OpenAI, and any SDK that lets you set a base URL works with it. OpenRouter-only options such as 

Can I switch between Anthropic, OpenAI and Mistral with one API key?

Yes. OpenRouter and Vercel AI Gateway bill Claude, GPT and Mistral models against one key and one prepaid balance. Cloudflare AI Gateway does the same for Claude and GPT, but not for Mistral's API. Portkey and LiteLLM also give you one gateway key, but you store your own vendor keys behind it.

Should the base URL include /chat/completions?

No. The OpenAI SDKs and BrowseWiz both append 

 to the base URL. Including it yourself doubles the path, and OpenRouter answers with a 404.

Partly. Anthropic offers an OpenAI SDK compatibility layer at 

 and strict tool schemas, and describes the layer as meant for testing and comparing models. OpenRouter lists 

 as supported parameters for Claude 4.5 and later models, so it is the simpler path if you need those through the OpenAI format.

OpenRouter API base URL and OpenAI-compatible endpoints

OpenRouter's base URL is https://openrouter.ai/api/v1. Copy-paste curl and Python code, plus other ways to switch between Anthropic, OpenAI and Mistral.

OpenRouter API Base URL & OpenAI-Compatible Endpoints

OpenRouter API Base URL and OpenAI-Compatible APIs: One Endpoint for Claude, GPT and Mistral

This n8n invoice processing workflow collects invoices and receipts from Gmail for a date range you choose. It downloads every PDF attachment, asks an OpenAI model whether each file is an invoice or a receipt, uploads the matches to a new Google Drive folder and, if you ask it to, emails the batch to your accountant. The flowchart below shows the path node by node, and because a Webhook node starts it, you can run it from a BrowseWiz chat message, curl or an NFC tag.

 in April 2025. This version uses the node names and settings n8n shows as of September 2026 and fixes three bugs in the original.

The n8n invoice processing workflow as a flowchart

The table adds the helper nodes the diagram leaves out.

How to build the invoice processing workflow in n8n

Import the template and add your Gmail, Google Drive and OpenAI credentials, or build it by hand with these settings.

The workflow answers early because a month of email can take minutes, and on n8n Cloud a webhook that doesn't respond within 100 seconds fails with a 524 error. n8n runs parallel branches from the top of the canvas down, so keep Respond to Webhook above the Configure branch or the reply waits for the whole run.

An Edit Fields (Set) node named Configure holds the settings the other nodes read: 

 from a chat tool never sent the email. This expression accepts both:

const results = [];
for (const item of $input.all()) {
  for (const key of Object.keys(item.binary ?? {})) {
    results.push({ json: {}, binary: { data: item.binary[key] } });
  }
}
return results;

Does this PDF file look like a {{ $('Configure').first().json['Match on'] }}?
Return "true" if it is and "false" if not. Reply only with lowercase true or false.
Filename: {{ $binary.data.fileName }}
Text: {{ $json.text }}

const merged = { json: {}, binary: {} };
$input.all().forEach((item, i) => {
  merged.binary[`file_${i}`] = item.binary.data;
});
return [merged];

Then publish the workflow (Publish button or Shift+P). The production URL only runs the published version; the test URL listens for 120 seconds after you select Listen for test event.

Trigger the n8n workflow from BrowseWiz in natural language

BrowseWiz is an AI side panel for Chrome and Edge. Its custom tools send a JSON body that the model fills in from your message to an endpoint you register, and an n8n Webhook accepts that body as it is.

Check that the tool is switched on under Custom Tools in the wrench menu, then ask something like "Collect my invoices and receipts from August 2026 and email them to my accountant." The model turns that into a request body:

{
  "startDate": "2026-08-01T00:00:00Z",
  "endDate": "2026-09-01T00:00:00Z",
  "sendEmail": true
}

The webhook's reply goes back to the model, so the chat answer can include the Drive folder link. The rest of the run shows up in n8n's Executions tab.

The model decides by itself when to call a tool, and BrowseWiz's docs advise caution with endpoints that change data, so switch this one off when you don't need it. If it picks the wrong range for "last month", give it the dates.

Trigger an n8n workflow from an NFC tag, phone shortcut or curl

n8n has no NFC trigger node, so the tag has to make your phone call a webhook. A phone can't easily add the auth header the main webhook expects, so I give it a small workflow of its own:

Then point the phone at the helper webhook's URL:

curl -X POST "https://your-n8n.example.com/webhook/<path>" \
  -H "X-Invoice-Key: <your-secret>" \
  -H "Content-Type: application/json" \
  -d '{"startDate":"2026-08-01T00:00:00Z","endDate":"2026-09-01T00:00:00Z","sendEmail":false}'

Variations: scanned receipts, separate folders, long files

How do I see an n8n workflow as a flowchart?

The n8n editor canvas already works as one. Nodes are steps, connections are arrows, and If nodes split into true and false outputs. Press 1 to zoom to fit, Shift+S to add a sticky note, and on n8n 2.31.0 or later Ctrl/Cmd+G to put related nodes in a named Canvas Group. For a diagram outside n8n, download the workflow JSON from the editor menu and write the node names as a Mermaid 

Yes. This workflow runs on demand through its webhook, but the same chain can start from a Schedule Trigger on the first of each month (with the 

 date expressions above) or from a Gmail Trigger for each new email. Spot-check what the AI picks.

How do I trigger an n8n workflow from an NFC tag?

n8n has no NFC trigger node, so the tag has to make your phone call a Webhook URL. On iPhone, a Shortcuts automation with the NFC trigger runs a Get Contents of URL action. On Android, a tag holding the URL opens it as a GET request, so the Webhook node must accept GET.

Extract From File with the Extract From PDF operation. Read PDF is on n8n's deprecated list, so it still runs but gets no fixes. When you swap it, set Keep Source to Both or the PDF is dropped from the item.

n8n invoice processing workflow flowchart

Node-by-node flowchart of an n8n invoice processing workflow: Gmail PDFs checked by OpenAI, saved to Google Drive, started from chat or an NFC tag.

n8n Flowchart Example: Invoice Processing Workflow

n8n Invoice Processing Workflow (with Flowchart): Collect Invoices and Receipts from Gmail

As of September 2026, the frontier AI models are OpenAI's GPT-6 Astra, Anthropic's Claude Opus 5.5 and Claude Fable 5.1, Google's Gemini 3.1 Pro and Gemini 3.8 Flash, xAI's Grok 4.7 and Meta's Muse Spark 1.3. The strongest open-weight models come from Chinese labs: Xiaomi's MiMo-V2.6-Pro, Z.ai's GLM-5.3, Moonshot's Kimi K3, Alibaba's Qwen3.8 and DeepSeek's V4.1-Flash. Independent leaderboards split the top spot, with GPT-6 Astra leading the hardest math, science and reasoning tests and Claude Opus 5.5 leading the Artificial Analysis Intelligence Index and Arena's human-vote ranking.

This page replaces our March 2025 comparison of GPT-4o, Claude 3.7 Sonnet and Gemini 2.0. Specs and prices come from vendor docs, checked on September 27, 2026. If you are looking for the 2024 or 2025 lineup, the timeline after the lab-by-lab section lists the frontier releases since GPT-4o.

Current top frontier AI models compared (September 2026)

Prices are for standard-length prompts; OpenAI charges more above 272K input tokens, Google (3.1 Pro) and xAI above 200K. Gemini 3.8 Flash's price is introductory until December 31, 2026 ($1.50 / $7.50 after). DeepSeek's input price is the cache-miss rate, and both DeepSeek ranges are off-peak / peak.

 on September 3, 2026 as its most capable model, "built for the hardest end-to-end work". It went to a limited set of organizations first and, over the following days, to ChatGPT Plus, Pro, Business and Enterprise users and the API. Astra is OpenAI's 

 at the "Critical" level of cybersecurity capability in its Preparedness Framework, so at launch it refuses advanced security tasks such as creating proof-of-concept exploits. 

 followed on September 22, Sol for coding and agents and Luna for high-volume work at $0.10 / $0.50. All three take text and images, with a 1.05M-token context window and up to 128K tokens of output (

Anthropic: Claude Opus 5.5 and Claude Fable 5.1

Anthropic's current lineup is Fable 5.1, Opus 5.5, Sonnet 5 and Haiku 4.5 (

, released September 22, 2026, is the one Anthropic recommends starting with. Anthropic says it "performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5", and it lists at $4 / $20 (

, released September 1, is the pricier option, which the models overview positions for "demanding reasoning and long-horizon agentic work". Both have a 1M-token context window at standard pricing, 128K output and a June 2026 knowledge cutoff. Claude Mythos 5.1, the same model with safeguards built for cybersecurity and life-sciences work, is invitation-only.

Google DeepMind: Gemini 3.1 Pro and Gemini 3.8 Flash

, in preview since February 19, 2026. Google said in May that Gemini 3.5 Pro would roll out 

. The Flash line moved almost monthly instead, reaching 

 now points new projects to it. Gemini 3.1 Pro shares second place on Humanity's Last Exam and 3.8 Flash is level with GPT-6 Astra on GPQA Diamond, but on the broader Artificial Analysis index they score 30 and 41, below the best open-weight models. Gemini models accept text, images, video, audio and PDFs, with a 1M-token window and 64K output (

xAI, which now brands itself SpaceXAI, released 

 on September 21, 2026, about six weeks after Grok 4.6. It targets "coding, agentic tasks, and knowledge work", takes text and images, and has a 500K-token context window (

). It is on the xAI API, in Cursor and on gateways such as OpenRouter; the 

 says the consumer Grok apps get it later.

Meta's frontier model is no longer a Llama. 

 launched on April 8, 2026 from Meta Superintelligence Labs without downloadable weights. 

 shipped on September 2 with better agentic and coding performance, and it is available through the Meta Model API (

). At its max reasoning setting it is the highest-scoring model on the Artificial Analysis index outside Anthropic and OpenAI. Meta says an open-weight Muse Spark release is on its roadmap. For now Meta's newest open-weight language model is the much smaller 

 (30B, Apache 2.0); the last large Llama was 

 as a preview on April 24, 2026, with MIT-licensed weights. Its API now runs 

, released September 10 with open weights, and V4-Pro is being phased out: since September 14 its requests are served by V4.1-Flash until a V4.1-Pro launches.

Other labs now sit ahead of DeepSeek among open-weight models on the 

. Xiaomi's MiMo-V2.6-Pro leads that list at 46 points, against 58 for Claude Opus 5.5, followed by Z.ai's 

 (July, 2.8 trillion parameters) at 44. Alibaba's 

 (August) takes images and video in the API, while its open-weight version, Qwen3.8-2.4T-A95B, is text-only and scores level with DeepSeek V4.1-Flash. MiMo and DeepSeek weights are MIT-licensed; Kimi K3, GLM-5.3 and Qwen3.8 use custom licenses that add conditions for model-hosting businesses above a revenue threshold.

Frontier AI models in 2024 and 2025: what changed

Reasoning moved into the flagship. In 2024 OpenAI's o1 was a separate "reasoning model"; GPT-6 Astra and Grok 4.7 now expose a reasoning-effort setting on the main model, and thinking is on by default for Claude 5 models (

A 1M-token window became normal. Gemini 1.5 Pro launched in 

 with a standard 128K window and 1M only in private preview; every model in the table except Grok 4.7 now takes about a million tokens.

What are frontier models? Frontier AI vs generative AI

"Frontier" is a moving label for the most capable general-purpose models of the moment. The 

 says frontier AI "refers broadly to those general purpose AI models that constitute the state of the art, a collection which will shift over time as the field progresses." The UK government's version, written for the 2023 AI Safety Summit, is "highly capable general-purpose AI models that can perform a wide variety of tasks and match or exceed the capabilities present in today's most advanced models" (

Lawmakers count training compute instead. California's 

, in force since January 1, 2026, defines a frontier model as a foundation model trained with more than 10²⁶ operations. The EU AI Act presumes that a general-purpose model trained with more than 10²⁵ FLOPs has "high impact capabilities", the trigger for its systemic-risk rules (

Generative AI is the wider category: any model that produces text, images, audio, video or code. Every frontier language model is generative AI, but a small open model running on a laptop is generative too and nowhere near the frontier. "Frontier" says where a model sits on the capability curve; "generative" says what it does.

Frontier AI benchmarks: who leads FrontierMath, GPQA, HLE and ARC-AGI

Labs run benchmarks with their own settings, so the numbers here come only from leaderboards the benchmark owners publish (checked September 27, 2026). Claude Opus 5.5 came out on September 22 and is still missing from most of them.

FrontierMath (Epoch AI, not Google DeepMind)

 is a set of very hard math problems: Tiers 1-3 run from undergraduate to advanced graduate level, and Tier 4 is research-level mathematics. Epoch AI runs it with funding from OpenAI, and Epoch 

 that OpenAI commissioned the core problems and has access to much of the set, apart from a holdout.

, GPT-6 Astra leads Tiers 1-3 at 93.7%, with Claude Fable 5.1 second at 90.2%. Astra also leads Tier 4 at 97.6%. That is higher than its Tiers 1-3 score because the private Tier 4 set has only 41 problems, and Astra solved 40 of them. Epoch's best Gemini result on Tiers 1-3 is 3.7 Flash at 71.6%.

Google DeepMind doesn't run FrontierMath, but it has commissioned and published results on it. In October 2025 Google commissioned Epoch to test Gemini 2.5 Deep Think, which set records at the time with 

 for its AI co-mathematician, an agent built on Gemini 3.1 Pro that Epoch tested blind. Google's Gemini model pages don't list FrontierMath scores.

 is 198 multiple-choice questions in biology, physics and chemistry, written by domain experts to be "Google-proof". On Epoch's runs GPT-6 Astra (95.8%), Gemini 3.8 Flash (95.4%) and Gemini 3.7 Flash (94.8%) top the list, but about ten models sit within the leader's margin of error, so the test now mostly confirms a model is top-tier.

 gives a model a codebase and a real GitHub issue to fix. The official 

 leaderboard (500 human-checked tasks) hasn't added a model since February 2026, so its bash-only leader, Claude Opus 4.5 at 76.8%, is several generations old. Scale AI's 

 board stops at July 2026 and lists none of the September models. For the newest coding models you only have vendor numbers, each run on the vendor's own agent setup.

, from the Center for AI Safety and Scale AI, has 2,500 questions from nearly 1,000 subject experts. On 

, GPT-6 Astra is first at 54.8%. Claude Fable 5.1 (46.5%), Gemini 3.1 Pro (46.4%), Gemini 3.8 Flash (44.5%) and GPT-5.4 Pro (44.3%) share second place, and Scale marks every entry with a potential-contamination warning.

, where ARC Prize tests models on its own semi-private set, GPT-6 Astra leads ARC-AGI-2 at 95.0%, and Claude Opus 5.5 reaches 93.3% at $0.41 per task against Astra's $1.12. ARC-AGI-2 is close to saturated (the human panel scores 100%). On the newer ARC-AGI-3, Astra leads at 62.7% on the standard harness, and the best Anthropic entry, Claude Opus 5, scores 30.2%.

Two broader rankings put Anthropic first. The 

 (v4.3) scores Opus 5.5 at 58, ahead of Fable 5.1 and GPT-6 Astra at 53. 

 (formerly LMArena), built from human votes, ranks Opus 5.5 first, within the margin of error of several other Claude models; GPT-6 Astra is 26th.

How to pick a frontier model for your task

Because the leaderboards disagree at the top, test two or three candidates on a real task of yours before committing.

Trying several frontier models side by side in BrowseWiz

BrowseWiz is an AI side panel for Chrome and Edge that chats with a model about the page, PDF or selection in front of you. It offers 

Most labs in this post document such an endpoint:

 as intended "to test and compare model capabilities", which is this exact job. Our 

 walks through the setup, OpenRouter included.

To compare, attach the page or file you're working on and ask your question. Then pick another model in the model menu and click 

 on the answer. BrowseWiz regenerates it with the newly selected model and the original context, so every model sees the same input. Each answer is kept, and arrows on the message (1/2, 2/2) switch between them. Do this on the latest answer, because Try again removes any messages that came after it.

As of September 2026: GPT-6 Astra (OpenAI), Claude Opus 5.5 and Fable 5.1 (Anthropic), Gemini 3.1 Pro and 3.8 Flash (Google), Grok 4.7 (xAI) and Muse Spark 1.3 (Meta). MiMo-V2.6-Pro, GLM-5.3, Kimi K3, Qwen3.8 and DeepSeek V4.1-Flash lead the open-weight models.

No single model wins everywhere. GPT-6 Astra leads FrontierMath, Humanity's Last Exam and ARC-AGI. Claude Opus 5.5 leads the Artificial Analysis Intelligence Index, ranks first on Arena's text leaderboard and costs 40% of Astra's per-token price.

What is the difference between frontier AI and generative AI?

Generative AI is any model that creates text, images, audio, video or code. Frontier AI is the small set of general-purpose models at the state of the art, a set the Frontier Model Forum says "will shift over time". Frontier language models are generative; most generative models aren't frontier.

Is FrontierMath a Google DeepMind benchmark?

No. Epoch AI runs FrontierMath, with funding from OpenAI. Google DeepMind has reported results on it, most recently 48% on Tier 4 for its AI co-mathematician agent in May 2026. GPT-6 Astra now leads both Tiers 1-3 (93.7%) and Tier 4 (97.6%) on Epoch's leaderboard.

What were the frontier AI models in 2024 and 2025?

In 2024: GPT-4o, Claude 3.5 Sonnet, Gemini 1.5 Pro and Llama 3.1 405B, with OpenAI's o1 arriving as a preview in September. At the start of 2025: GPT-4o and o1, Claude 3.5 Sonnet, Gemini 2.0 Flash and DeepSeek-V3, joined by DeepSeek-R1 in January. By December 2025: GPT-5.2, Claude Opus 4.5, Gemini 3 Pro, Grok 4.1 and DeepSeek V3.2.

Frontier AI Models in 2026: The Current Lineup Compared

GPT-6 Astra or Claude Opus 5.5? The current frontier AI models (September 2026) compared on price and benchmarks, plus what changed since 2024 and 2025.

Learn more about AI and BrowseWiz

OpenRouter API Base URL and OpenAI-Compatible APIs: One Endpoint for Claude, GPT and Mistral

n8n Invoice Processing Workflow (with Flowchart): Collect Invoices and Receipts from Gmail

Current Frontier AI Models (September 2026): GPT-6, Claude, Gemini and Grok Compared