# Mesh API

> One API. Every AI Model. An OpenAI-compatible gateway to 400+ LLMs with automatic fallbacks, unified billing, and INR/UPI payments for Indian teams.

This file is a comprehensive plain-text mirror of https://meshapi.ai for AI agents and crawlers. It captures the site's product, models, pricing, security posture, research, FAQs, and policies in a single fetch.

Site: https://meshapi.ai
Docs: https://developers.meshapi.ai/
API base URL: https://api.meshapi.ai/v1
Status: https://status.meshapi.ai
Company: Fiesta Labs Inc., 1 Chestnut Hill Plaza #1088, Newark, DE 19713, USA
Contact: contact@meshapi.ai

---

## About

Mesh API is a unified inference gateway. Developers swap their OpenAI base URL and key for Mesh and instantly route requests across 400+ models from OpenAI, Anthropic, Google, DeepSeek, Meta, Mistral, xAI, Qwen, Moonshot, Z.ai, MiniMax, NVIDIA, and more. The gateway adds automatic fallbacks, load balancing, cost/latency observability, and native INR billing with UPI top-ups for Indian teams.

Mesh API is a product of Fiesta Labs Inc.

## Product Overview

- OpenAI-compatible endpoint: point any OpenAI SDK at `https://api.meshapi.ai/v1` and swap the API key. No code changes.
- 400+ models across every major provider, exposed through one schema.
- Intelligent routing with automatic failover to a backup model in under 100ms if a provider degrades.
- Load balancing across upstream providers and regions.
- Unified billing in USD or INR. UPI, cards, and enterprise invoicing supported.
- Sub-80ms worst-case overhead; typically 20-40ms.
- 99.9% uptime SLA on pay-as-you-go; custom SLAs on volume contracts.
- Prompt caching and token-savings dials (see Mesh Optimize research).
- India-first: INR pricing, UPI, deployable from India, GST invoicing.

## Featured Models (curated leaderboard)

Ranked by internal intelligence index. Full 400+ catalog is available to authenticated users in the dashboard.

| Rank | Provider     | Model                     | Context  | Intelligence |
|------|--------------|---------------------------|----------|--------------|
| 1    | Anthropic    | Claude Fable 5            | 1M       | 60           |
| 2    | Anthropic    | Claude Opus 4.8           | 1M       | 56           |
| 3    | OpenAI       | GPT-5.5                   | 1.05M    | 55           |
| 4    | Anthropic    | Claude Sonnet 5           | 1M       | 53           |
| 5    | Z.ai         | GLM-5.2                   | 1.05M    | 51           |
| 6    | Google       | Gemini 3.5 Flash          | 1M       | 50           |
| 7    | Google       | Gemini 3.1 Pro Preview    | 1.05M    | 46           |
| 8    | Qwen         | Qwen3.7 Max               | 1M       | 46           |
| 9    | MiniMax      | MiniMax-M3                | 1.05M    | 44           |
| 10   | DeepSeek     | DeepSeek V4 Pro           | 1.05M    | 44           |
| 11   | Moonshot AI  | Kimi K2.6                 | 262K     | 43           |
| 12   | Xiaomi       | MiMo-V2.5-Pro             | 1.05M    | 42           |
| 13   | DeepSeek     | DeepSeek V4 Flash         | 1.05M    | 40           |
| 14   | OpenAI       | GPT-5.4 mini              | 400K     | 40           |
| 15   | NVIDIA       | Nemotron 3 Ultra          | 512K     | 38           |
| 16   | xAI          | Grok 4.3                  | 1M       | 38           |
| 17   | Qwen         | Qwen3.5 397B A17B         | 262K     | 34           |
| 18   | Mistral AI   | Mistral Medium 3.5        | 131K     | 30           |
| 19   | Anthropic    | Claude 4.5 Haiku          | 200K     | 30           |
| 20   | Google       | Gemma 4 31B               | 262K     | 29           |
| 21   | OpenAI       | gpt-oss-120b              | 128K     | 24           |
| 22   | OpenAI       | gpt-oss-20b               | 128K     | 15           |
| 23   | Upstage      | Solar Pro 3               | 128K     | 14           |

## Pricing

- Pay-as-you-go: no minimums, no monthly commitment, per-token billing at provider rates plus a small routing margin.
- Team and Enterprise: volume discounts, dedicated capacity, custom SLAs, invoicing, and custom contracts (which supersede standard terms where applicable).
- INR top-ups via UPI available for Indian customers.
- Full per-model pricing at https://meshapi.ai/pricing
- Methodology: https://meshapi.ai/pricing-methodology
- Auto Routing: https://meshapi.ai/auto-routing

All payments are non-refundable. See /terms for the full policy.

## Migration Guides

Drop-in migration from other gateways with two-line config changes:

- LiteLLM -> Mesh: https://meshapi.ai/migration/litellm
- OpenRouter -> Mesh: https://meshapi.ai/migration/openrouter
- Vercel AI Gateway -> Mesh: https://meshapi.ai/migration/vercel-ai-gateway
- Overview: https://meshapi.ai/migration

## Security & Compliance

Status page: https://meshapi.ai/security

Assessments in progress (initiated May 2026 with Scrut Automation, 8-10 week timeline):
- SOC 2 Type 2
- ISO 27001
- EU GDPR

Data handling (default posture):
- Zero Data Retention (ZDR) is ENABLED BY DEFAULT on every account. Prompt and completion content is processed in memory and never written to Mesh logs or databases.
- Mesh never uses customer prompts, completions, files, or embeddings to train, fine-tune, or evaluate any model.
- Mesh never sells, rents, or shares customer content for advertising or marketing.
- Only usage metadata is retained: model, token counts, latency, cost, request ID. No prompt or completion text.
- If a customer explicitly disables ZDR, prompt/completion content and related context may be retained for up to 90 days, and longer only where law or a signed contract requires it. The no-training and no-selling commitments still apply.
- Content routed to a third-party model provider is subject to that provider's own terms and retention practices.

Security features:
- Prompt injection prevention
- DDoS protection
- Load balancing across providers and regions
- 100% traceability of every request
- PII protection

Baseline: TLS 1.2+ in transit, encryption at rest, least-privilege access, audit logging. Security contact: contact@meshapi.ai

## Research

Published papers and benchmarks at https://meshapi.ai/research:

- Dollar per Image Bench (https://meshapi.ai/research/dollar-per-image-bench): 7 frontier image models on Mesh. Cheapest image beats the most expensive at 1/7.5th the price.
- Dollar per Task Bench (https://meshapi.ai/research/dollar-per-task-bench): 5 LLMs across 10 real tasks. Cheapest model passed every coding test with a 259x bill spread.
- Value Rankings (https://meshapi.ai/research/value-rankings): capability per dollar across the full text-model catalog. 122 verified rankings + full price sheet.
- Mesh Optimize (https://meshapi.ai/research/mesh-optimize): engineering notes on the token-savings dial. Up to ~95% input savings on cached prefixes.

## FAQ

Q: Is Mesh compatible with the OpenAI Python and Node.js SDK?
A: Yes. Set base_url to https://api.meshapi.ai/v1 and swap your API key for your Mesh key. Every method, parameter, and response format is identical. No code changes beyond those two values.

Q: How much latency does Mesh add to my requests?
A: Under 80ms in the worst case. Most requests add 20-40ms of overhead. Our infrastructure is colocated close to the major provider endpoints to keep this as low as possible.

Q: What happens if a provider goes down mid-request?
A: Requests automatically fail over to your configured backup model within 100ms. You set fallback preferences in the dashboard or via the API. Your users see no interruption.

Q: Is my request content stored or logged?
A: No. Zero Data Retention is on by default for every account, so prompts and completions are processed in memory and never written to our logs. We log metadata only (model used, token count, cost, latency) for billing and analytics, and we never train on your data or sell it. Full details in our privacy policy.

Q: Can I use Mesh for production traffic at scale?
A: Yes. We offer a 99.9% uptime SLA for pay-as-you-go accounts and custom SLAs for volume agreements. Live status is available at status.meshapi.ai.

## Blog

Long-form posts at https://meshapi.ai/blog:

- Your Model Eval Framework Is Broken. Claude Detects Prefilled Prompts and Silently Resists. (/blog/prefill-awareness-evals-invalid)
- Google's Open Knowledge Format Is Smart. But Here's What's Missing. (/blog/introducing-the-open-knowledge-format-as-something-new-launched)
- Why 4B Models Beat 671B Giants at Theorem Proving. (/blog/pythagoras-prover-small-models-reasoning)
- Google Analytics Now Tracks ChatGPT, Claude, and Gemini Traffic Separately. (/blog/google-analytics-ai-assistant-traffic)
- AI Agents Fail Real Workflows at Scale. (/blog/agents-last-exam-production-gap)
- Stop Fine-Tuning Tool Workflows. Evolutionary Search Raises Small Models From 3% to 24%. (/blog/evoflux-tool-workflows-beats-finetuning)
- 60-80% of Agent Traffic Goes Local Now. (/blog/routing-local-first-60-80-pattern-default)
- LangChain and AutoGPT Have Zero Memory-Integrity Safety. (/blog/containment-gap-langchain-agent-safety)
- When Prompt Caching Actually Breaks Even. (/blog/prompt-caching-latency-roi-threshold)
- Why Your LLM Provider's Generous Daily Quota Is a Lie. (/blog/llm-api-tpm-ceiling-kills-daily-quota)
- Your Agent Task Costs $0.50 in Tokens but $8 in Retries. (/blog/agent-workflows-real-cost-retries-vs-tokens)
- We Made Two LLMs Play Tetris. (/blog/llm-tetris-contest-opus-deepseek)
- Why Claude Design Burns Your Quota (and How to Fix It). (/blog/why-claude-design-burns-quota)

## Community

Projects, demos, and experiments built on Mesh API by the developer community: https://meshapi.ai/community

## Careers

We are hiring across routing infrastructure, billing rails, and developer experience. Open roles and application form at https://meshapi.ai/careers

## Legal

- Privacy Policy: https://meshapi.ai/privacy
- Terms & Conditions: https://meshapi.ai/terms (pay-as-you-go; custom Enterprise/Team contracts take precedence; all payments non-refundable)
- Security & Compliance: https://meshapi.ai/security

## Full Page Index

- / — Home. Product overview, featured models, live usage stats.
- /pricing — Plans and per-model rates.
- /pricing-methodology — How Mesh pricing is calculated.
- /auto-routing — Technical report on intelligent auto routing: reliability plane (retries, fallbacks, latency budget), performance plane (pinned / auto / smart modes), per-task efficiency frontiers, observability, governance, and deployment models.
- /careers — Open roles.
- /migration — Migration overview.
- /migration/litellm — Move from LiteLLM.
- /migration/openrouter — Move from OpenRouter.
- /migration/vercel-ai-gateway — Move from Vercel AI Gateway.
- /research — Research index.
- /research/dollar-per-image-bench — Image model cost/quality benchmark.
- /research/dollar-per-task-bench — Text model cost/quality benchmark.
- /research/value-rankings — Full price-performance leaderboard.
- /research/mesh-optimize — Token savings engineering notes.
- /blog — Engineering and research blog.
- /community — Community-built projects.
- /security — Security features and compliance status.
- /privacy — Privacy policy.
- /terms — Terms & conditions.

## Developer Resources

- API reference and SDK guides: https://developers.meshapi.ai/
- OpenAI SDK compatibility: set `base_url = "https://api.meshapi.ai/v1"` and use your Mesh API key.
- Status and incident history: https://status.meshapi.ai
- Support: contact@meshapi.ai