The unified gateway for web search

Q: How does pricing work?

You buy credits once and spend them on any provider. Most providers bill per request rather than per result, so a 10-result search costs the same as a 5-result one, and those rates run from $4 to $8 per 1,000 searches. Two providers scale with the result count instead: Exa is flat at $7 through 10 results and then adds $1 per 1,000 for each extra result, reaching $17 at 20, and Valyu bills every result it returns at up to $1.50 per 1,000, so a 1-result search is $1.50 and a 20-result one is $30. Every response returns its own usage.cost, so you can attribute spend request by request.

Q: Is there a platform fee?

Yes, 5% on credit purchases. Top up $100 and you are charged $105, with the full $100 landing in your balance to spend. Nothing is added on top of a search itself: every request draws down your balance at the provider rate listed above, so the cost you see on a response is the cost you pay.

Q: What's the difference between OpenWebSearch and going direct to each provider?

Going direct means a separate contract, key, client, and response shape for every index you want to use. Here you integrate once: one key, one request format, and one normalized result shape across providers. Pass an ordered providers list and a request falls through to the next index when one fails or degrades, and switching providers is a change to a single request field rather than an integration rewrite. The untouched provider payload is still returned under raw whenever you need something provider-specific.

Q: How does it work with my existing AI stack?

It is a plain REST endpoint that takes and returns JSON, so it drops into any language or framework and wires up as a tool or function call like any other API. Results arrive in a normalized shape with title, url, snippet, and content, which is the form most RAG and agent pipelines already expect, so you do not need a per-provider adapter. GET /v1/providers reports the live providers and their limits at runtime instead of you hardcoding them.

Q: Does it support zero data retention routing?

Yes for your search content. Query text and the results we return are not retained; they pass through the request and are gone. We keep only the request metadata needed to bill and show usage, which is the request id, the provider that served it, the cost, and the latency. One caveat worth knowing: whichever provider a request routes to applies its own retention policy to the query it receives, so if you need a specific guarantee, check the policy of the providers you route to and pin your provider list accordingly.

All search indexes, one API key, one structured format and high availability.

Exa

Tavily

Brave

Bing

Perplexity

Interfaze

Parallel

Apify Serp

Valyu

Octen

+ more

TRY

Live demo unavailable: NEXT_PUBLIC_TURNSTILE_SITE_KEY is not set

Results

JSON

464 ms · $0.008

Best Open Source LLMs 2026 - Telnyx

telnyx.com

DeepSeek V4, GLM 5.2, Kimi K2.6, Kimi K3, MiniMax M3, Qwen3 VL, and Gemma 4. Compare the best open-source LLMs of 2026 by benchmarks, context window, and use case. ... - The best open-source LLMs in 2026 are DeepSeek V4, GLM 5.2, Kimi K2.6, Kimi K3, MiniMax M3, Qwen3 VL 235B, and Google Gemma 4 31B. - Kimi K3 posts the highest GPQA Diamond score of any open model at 93.5%, per the Onyx leaderboard, an independent benchmark aggregator, with full weights released July 27, 2026. - The latest open source LLMs now match or beat proprietary models on reasoning and coding benchmarks. Mixture-of-experts architecture is the default across the field. ... ## The best open-source LLMs in 2026 The best open-source LLMs in 2026 are DeepSeek V4, GLM 5.2, Kimi K2.6, Kimi K3, MiniMax M3, Qwen3 VL 235B, and Google Gemma 4 31B. These seven lead the latest open source LLMs on benchmarks and real-world use. Kimi K3, with full weights released July 27, posts 93.5% on GPQA Diamond. ... OpenAI also entered the open-source space with GPT oss 120b and 20b, though those models are not yet frontier-grade. ... ## Comparing the best open-source LLMs |Model|Context window|Standout benchmark| |--|--|--| |DeepSeek V4 Pro|1M|80.6% SWE-Bench| |DeepSeek V4 Flash|1M|85.9% BrowseComp| |GLM 5.2|1M|54.7% Humanity's Last Exam| |Kimi K2.6|256K|80.2% SWE-Bench| |Kimi K3|1M|93.5% GPQA Diamond| |MiniMax M3|1M|93% GPQA Diamond| |Qwen3 VL 235B|256K (up to 1M)|87.1% MMLU| |Gemma 4 31B|262K|85.2% MMLU|

Open Source LLM Leaderboard 2026 - Vellum

vellum.ai

updated 24 Jul 2026 # Open Source LLM Leaderboard This open source LLM leaderboard displays the latest public benchmark performance for open-weight and open-source models released after April 2024. The data comes from model providers as well as independently run evaluations by Vellum or the open-source community. We feature results from non-saturated benchmarks, excluding outdated benchmarks (e.g. MMLU). ... 56 Kimi K3 54.7 GLM 5.2 54 Kimi K2.6 51.6 DeepSeek V4 Flash 48.2 DeepSeek V4 Pro 44.9 Kimi K2 Thinking 30.1 Kimi K2.5 14.9 GPT oss 120b 10.9 GPT oss 20b 8.6 DeepSeek-R1 ... |Kimi K3|56%| |GLM 5.2|54.7%| |Kimi K2.6|54%| |DeepSeek V4 Flash|51.6%| |DeepSeek V4 Pro|48.2%| |Kimi K2 Thinking|44.9%| |Kimi K2.5|30.1%| |GPT oss 120b|14.9%| |GPT oss 20b|10.9%| |DeepSeek-R1|8.6%| ... |Kimi K2.5|76.8%| New ... |MiniMax M3|70.1%| ... |Model|Context size|Cutoff date|I/O cost|Max output|Latency|Speed| ... |Kimi K3|1,048,576|-|$3 / $15|-|4.46s|35.2 t/s| |GLM 5.2|1,000,000|Mar 2026|$0.95 / $3|128,000|1.14s|347 t/s| ... |Models|Context Window|Input Cost / 1M tokens|Output Cost / 1M tokens|Speed (tokens/second)|Latency|

Best Open Source LLMs in 2026: We Reviewed 7 Models

fireworks.ai

Best Open-Source LLMs (Updated July 2026): Top Models - AceCloud

acecloud.ai

As of July 2026, GLM-5.2 is the strongest all-round open-weight LLM in this comparison, while Kimi K2.7 Code stands out for coding agents, Gemma 4 12B is a practical laptop model, and Nemotron 3 Super suits teams prioritizing open training resources. The right choice still depends on workload, hardware and licence. ... - **Best overall:** GLM-5.2, particularly for long-context coding, reasoning and agentic work. - **Best for coding agents:** Kimi K2.7 Code at data-centre scale; Qwen3-Coder-Next for a more efficient coding server. - **Best for local use:** Gemma 4 for laptops and edge devices; Qwen3.6-27B for higher-end 24GB systems. - **Best for enterprise:** Nemotron 3, because NVIDIA publishes weights, training data, recipes and evaluation resources. ... An **open-weight LLM** makes its trained parameters available for download. You can usually run it on your infrastructure, quantize it and fine-tune it, subject to its licence. A fully **open-source AI system**, under the Open Source Initiative’s definition, must also provide the freedoms to use, study, modify and share the system. That requires sufficient training-data information, training and inference code, and model parameters in the preferred form for modification. ... ## Best Open-Source LLM Models in 2026 |Model|Best use|Licence|Context|Deployment reality| |--|--|--|--|--| |GLM-5.2|Best overall; long-horizon coding and agents|MIT|1 million tokens|Very large server model; use multi-GPU infrastructure or an API.| |DeepSeek V4 Pro / Flash|Reasoning, knowledge work and efficient server inference|MIT|1 million|Pro has 1.6T total/49B active parameters; Flash has 284B/13B active. Both remain data-centre models.| |Kimi K2.7 Code|Long-running coding agents|Modified MIT|256K|1T total/32B active parameters; designed for distributed deployment rather than consumer GPUs.| |Qwen3.6-27B|High-end local use, coding and tool calling|Apache 2.0|262K native; extensible to about 1M|Practical in 4-bit form on some 24GB systems when context is reduced. Full-context official examples use eight GPUs.| |Gemma 4|Laptops, mobile devices and private edge applications|Apache 2.0|Up to 128K in the main edge stack|Includes E2B, E4B and 12B options, with extensive LiteRT support.| |Nemotron 3 Super / Ultra|Auditable agents, RAG and enterprise customization|NVIDIA Open Model Licence; newer releases moving to OpenMDW-1.1|1 million for Super|Publishes weights, datasets, recipes and evaluation resources; optimized most heavily for NVIDIA infrastructure.| |Mistral Small 4|Enterprise multilingual, multimodal and document workflows|Apache 2.0|256K|119B total/6B active; official minimum is four H100s, two H200s or one DGX B200.| ... For 2026, start with **GLM-5.2 for overall capability, DeepSeek V4 for reasoning, Kimi K2.7 Code for coding agents, Qwen3.6-27B for a high-end local system, Gemma 4 for laptops and edge devices, Nemotron 3 for reproducibility, and Mistral for multilingual enterprise applications**.

Best Open Source LLM Leaderboard 2026 - Onyx AI

onyx.app

Best Open Source Models — 2026 Rankings ... The definitive ranking of every major open source model — compared across reasoning, coding, math, software engineering, and instruction following benchmarks. Roshan Desai · Last updated: 2026-07-20 LLM Leaderboard (All Models) ... Kimi K3 2.8T GLM-5.2 753B DeepSeek-V4-Pro 1.6T Kimi K2.6 1T A MiniMax M3 428B DeepSeek-V4-Flash 284B Hunyuan Hy3 295B Step-3.7-Flash 198B Qwen3.6-27B 27B Qwen3.6-35B-A3B 35B Nemotron 3 Ultra 550B GLM-4.7 355B MiMo-V2-Flash 309B DeepSeek R1 671B B MiniMax M2.7 230B Gemma 4 31B 31B Mistral Medium 3.5 128B Command A+ 218B Ling-2.6-1T 1T Nemotron 3 Super 120B MiMo-V2.5-Pro 1.02T Mistral Small 4 119B GPT-oss 120B 117B Nemotron Ultra 253B 253B Nemotron Super 49B 49B Step3 316B C Llama 4 Maverick 400B Gemma 3 27B 27B Nemotron Nano 30B ... |Command A+Cohere|218B|128K|N/A|76.1|N/A|N/A|N/A|N/A|N/A|N/A|N/A|75.1| |DeepSeek R1DeepSeek|671B|128K|84.0|71.5|83.3|1398|49.2|90.2|65.9|87.5|97.3|90.8| ... |Gemma 3 27BGoogle|27B|128K|67.5|42.4|N/A|1366|N/A|N/A|29.7|N/A|89.0|N/A| |Gemma 4 31BGoogle|31B|262K|85.2|84.3|N/A|1451|N/A|N/A|80.0|N/A|N/A|N/A| |GLM-4.7Zhipu AI|355B|200K|84.3|85.7|88.0|1441|73.8|94.2|84.9|95.7|N/A|90.1| ... |GPT-oss 120BOpenAI|117B|128K|90.0|80.9|N/A|1355|62.4|88.3|60.0|97.9|N/A|90.0| |Hunyuan Hy3Tencent|295B|262K|N/A|90.4|N/A|1412|78.0|N/A|N/A|N/A|N/A|N/A| |Kimi K2.6Moonshot|1T|262K|N/A|90.5|N/A|N/A|80.2|N/A|89.6|N/A|N/A|N/A| |Kimi K3Moonshot|2.8T|1M|N/A|93.5|N/A|1486|N/A|N/A|N/A|N/A|N/A|N/A| |Ling-2.6-1TAnt Group|1T|262K|N/A|76.2|N/A|N/A|72.2|N/A|65.6|N/A|N/A|N/A| |Llama 4 MaverickMeta|400B|1M|80.5|69.8|N/A|1328|N/A|62.0|43.4|N/A|N/A|85.5| |MiMo-V2-FlashXiaomi|309B|262K|84.9|83.7|N/A|1393|73.4|84.8|80.6|94.1|N/A|86.7|

Citation-backed, real-time answers

normalized response

This is a limited preview demo. Create a free account for the full response, every provider, and control over each request.

Read the docs ->

Merge multiple providers

Query several providers in one request and get a single deduplicated, ranked list instead of stitching responses together yourself.

Fallbacks for high uptime

Give an ordered list of providers and we route to the next one automatically when one fails or degrades.

One response format

Receive a consistent result structure instead of handling a different payload for every provider.

Switch without rewrites

Test or change providers by updating one request field—not your application integration.

Switch search without switching code

Change the provider, not your integration. Requests and results stay consistent as providers evolve.

fetch

const response = await fetch("https://api.openwebsearch.ai/v1/search", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.OPENWEBSEARCH_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    provider: "exa",
    query: "What changed in browser automation this week?",
    max_results: 10,
  }),
});

const { results, provider, usage } = await response.json();

console.log(results);

Different indexes. Different strengths.

There is no universally best search provider. Historical depth, people data, freshness, media coverage, and ranking quality vary by index. Choose the provider that fits the job without changing your integration.

People & companies

Exa

Specialized indexes and structured fields for professional profiles and company discovery.

Specialized datasets

Interfaze

Valyu

Research, social, financial, academic, and proprietary sources beyond the general web.

Agent-ready context

Parallel

Tavily

Relevant excerpts and search modes designed to feed research agents and AI applications.

Broad web coverage

Brave

Bing

General web indexes with strong coverage across web pages, news, images, video, and local results.

Real-time answers

Perplexity

Octen

Fresh web context for grounded answers, live information, and multi-step research workflows.

SERP intelligence

Apify Serp

Localized organic results, ads, rich result blocks, related queries, products, and rankings.

Extract full webpage

Web scraping docs ->

Extract full webpage content in structured JSON or markdown using Interfaze AI to bypass bot detection for real-time web scraping.

linkedin.com/in/patrickcollison

{
  "firstName": "Patrick",
  "lastName": "Collison",
  "headline": "Stripe CEO",
  "openToWork": false,
  "hiring": false,
  "about": "Dynamically communicate prospective opportunities and proactive technologies. Efficiently aggregate interactive materials before state of the art collaboration and idea-sharing. Credibly supply cross-media metrics via leading-edge solutions.\n\nSeamlessly embrace world-class imperatives and technically sound best practices. Black belt parallelization of prospective relationships via tactical leadership skills. Enthusiastically productivate customer directed core competencies via impactful functionalities.\n\nUp-skilling internal or \"organic\" sources via innovative experiences. Quickly pursue enterprise-wide niche markets for ubiquitous best practices. Distinctively engineer process-centric markets before timely greenfield sources.",
  "premium": true,
  "influencer": true,
  "memorialized": false,
  "verified": true,
  "registeredAt": "2012-12-06T18:20:03.457Z",
  "connectionsCount": 500,
  "followerCount": 82886,
  "location": {
    "full": "San Francisco Bay Area",
    "countryCode": "US"
  },
  "experience": [
    {
      "companyName": "Arc Institute",
      "companyLinkedinUrl": "https://www.linkedin.com/company/arc-institute-org/",
      "position": "Cofounder",
      "location": "Palo Alto, California, United States",
      "employmentType": null,
      "workplaceType": null,
      "description": null,
      "skills": null,
      "startDate": "Jan 2021",
      "endDate": "Present"
    },
    {
      "companyName": "Stripe",
      "companyLinkedinUrl": "https://www.linkedin.com/company/stripe/",
      "position": "CEO",
      "location": null,
      "employmentType": null,
      "workplaceType": null,
      "description": null,
      "skills": null,
      "startDate": "2010",
      "endDate": "Present"
    },
    {
      "companyName": "Auctomatic",
      "companyLinkedinUrl": "https://www.linkedin.com/company/auctomatic/",
      "position": "Cofounder",
      "location": null,
      "employmentType": null,
      "workplaceType": null,
      "description": null,
      "skills": null,
      "startDate": "2007",
      "endDate": "2008"
    }
  ],
  "education": [
    {
      "schoolName": "Massachusetts Institute of Technology",
      "schoolLinkedinUrl": "https://www.linkedin.com/company/1503/",
      "degree": "Mathematics (incomplete)",
      "fieldOfStudy": null,
      "startDate": "2006",
      "endDate": "2010"
    }
  ],
  "volunteering": [],
  "courses": [],
  "skills": [
    "Programming Languages",
    "Software Engineering",
    "Artificial Intelligence",
    "Distributed Systems",
    "APIs",
    "Flying"
  ]
}

Structured output

Pricing by provider

Buy credits once and spend them on any provider below.

Parallel

Token-dense excerpts for AI agents

/ 1K searches

Brave

Independent broad-web and media search

/ 1K searches

Exa

People, companies, and semantic discovery

$7–$17

/ 1K searches

Perplexity

Citation-backed, real-time answers

/ 1K searches

Valyu

Academic, financial, and proprietary data

$1.50–$30

/ 1K searches

Tavily

Fresh news and agent-ready research

Coming soon

Bing

Broad web, local, news, and image coverage

Coming soon

Interfaze

Research, social, and financial indexes

Coming soon

Apify Serp

Localized SERP features and rankings

Coming soon

Octen

Real-time and multimodal web search

Coming soon

Rates measured live at 1 to 20 results. A range means the rate moves with the result count.

Pay only for the searches you run

One key and one prepaid balance across every provider.

Start searching in three steps

Start with one provider and switch whenever you need to. Your product keeps the same request and response contract either way.

Create a free account with Google. Add your team to the same org whenever you need to.

Continue with Google

Add credits

One prepaid balance that works across every provider. No separate contracts or invoices.

Apr 1

$99

Mar 30

$10

Get your API key

Create a key and send your first search. The same key works for every provider.

OPENWEBSEARCH_API_KEY

••••••••••••••••

FAQs

Read the docs ->

How does pricing work?

Is there a platform fee?

What's the difference between OpenWebSearch and going direct to each provider?

How does it work with my existing AI stack?

Does it support zero data retention routing?