Opper AI partners with Novita to bring 80+ open-weight models to the gateway. Stockholm, Sweden, July 2026. Opper AI and Novita are partnering to bring Novita's open-weight inference to the Opper AI gateway. Builders can now reach more than 80 Novita-hosted models through the Opper API, from DeepSeek V4 and Kimi K2.6 to GLM-5, MiniMax M3, Llama 4, and Qwen, making Novita one of the broadest single-provider catalogues on the gateway. Why this matters. One of the broadest open-weight catalogues anywhere. Novita serves models from close to twenty different makers through Opper: DeepSeek V4 Pro and Flash with 1M-token context windows, Moonshot's Kimi family up to K2.6 and the code-tuned K2.7, Z.ai's GLM-5, MiniMax M3 with a 1M-token window, Meta's Llama 3 and Llama 4, the Qwen lineup, Google's Gemma 4, OpenAI's GPT-OSS, Baidu's ERNIE 4.5, NVIDIA Nemotron, and a long tail of specialist checkpoints down to dedicated OCR models. When your agent needs one specific open-weight checkpoint, this is the route where it's most likely already live. New releases, production-ready fast. Open-weight releases move quickly, and Novita has made a name for being among the first providers to serve them with stable, production-grade inference. The current frontier of open models, Kimi K2.6, GLM-5, DeepSeek V4, and MiniMax M3, is already callable through Opper today, at price points that are consistently among the most competitive on the gateway. Per Novita's published figures, its model APIs run at around 200ms latency with 99.5% uptime, up to 50% cheaper than the large cloud providers. Clear data handling, US-hosted. Novita is hosted in the United States and does not train on customer data. The platform is SOC 2 Type II audited and a GDPR DPA is offered under standard contractual clauses; operationally, Novita keeps abuse-monitoring logs. For teams that want the breadth of the catalogue with a well-understood, audited posture behind it, that's a clean foundation to build on. More than a router: the AI control plane. Routing to Novita is just the entry point. Every call through Opper runs on its AI control plane: intelligent routing across models, full observability into every call, token, and session, real-time PII masking and content filtering, budget caps, and audit trails. Pin Novita for a task or set it as a fallback for rate limits and outages, and because Novita is OpenAI-compatible, getting there is a model string, not a migration. "We're glad to welcome Novita to Opper. They bring one of the broadest open-weight catalogues anywhere, from DeepSeek and Kimi to GLM, MiniMax, and Llama, and they're consistently among the first to get new releases production-ready. That gives our developers a deep bench of models they can reach the moment they need them." - Göran Sandahl, Co-founder and CEO, Opper AI "Developers come to Novita because new open models show up fast and just work in production. Opper puts that catalogue in front of a large community of builders through one API, with routing, observability, and governance handled, so teams can adopt new models the day they land." - Junyu Huang, COO, Novita Models live today. The catalog below is fetched live from Opper's model API and filtered to Novita-hosted models. Availability, context windows, and pricing stay in sync with what's actually callable through Opper. All Novita routes are hosted in the United States. | Model | Region | Context | Input / 1M | Output / 1M | | novita/Sao10K/L3-8B-Stheno-v3.2 | US | 8K | $0.05 | $0.05 | | novita/baichuan/baichuan-m2-32b | US | 131K | $0.07 | $0.07 | | novita/baidu/cobuddy | US | 131K | $0.28 | $1.13 | | novita/baidu/ernie-4.5-21B-a3b | US | 120K | $0.07 | $0.28 | | novita/baidu/ernie-4.5-21B-a3b-thinking | US | 131K | $0.07 | $0.28 | | novita/baidu/ernie-4.5-300b-a47b-paddle | US | 123K | $0.28 | $1.10 | | novita/baidu/ernie-4.5-vl-28b-a3b | US | 30K | $0.14 | $0.56 | | novita/baidu/ernie-4.5-vl-28b-a3b-thinking | US | 131K | $0.39 | $0.39 | | novita/baidu/ernie-4.5-vl-424b-a47b | US | 123K | $0.42 | $1.25 | | novita/deepseek-ocr | US | 128K | $0.03 | $0.03 | | novita/deepseek-ocr-2 | US | 128K | $0.03 | $0.03 | | novita/deepseek-v3.1 | US | 131K | $0.27 | $1.00 | | novita/deepseek-v3.2 | US | 164K | $0.27 | $0.40 | | novita/deepseek-v4-flash | US | 1.0M | $0.14 | $0.28 | | novita/deepseek-v4-pro | US | 1.0M | $1.74 | $3.48 | | novita/deepseek/deepseek-prover-v2-671b | US | 160K | $0.70 | $2.50 | | novita/deepseek/deepseek-r1 | US | 64K | $4.00 | $4.00 | | novita/deepseek/deepseek-r1-0528 | US | 164K | $0.70 | $2.50 | | novita/deepseek/deepseek-r1-0528-qwen3-8b | US | 128K | $0.06 | $0.09 | | novita/deepseek/deepseek-r1-distill-llama-70b | US | 8K | $0.80 | $0.80 | | novita/deepseek/deepseek-r1-distill-qwen-14b | US | 33K | $0.15 | $0.15 | | novita/deepseek/deepseek-r1-distill-qwen-32b | US | 64K | $0.30 | $0.30 | | novita/deepseek/deepseek-r1-turbo | US | 64K | $0.70 | $2.50 | | novita/deepseek/deepseek-r1/community | US | 64K | $4.00 | $4.00 | | novita/deepseek/deepseek-v3-0324 | US | 164K | $0.27 | $1.12 | | novita/deepseek/deepseek-v3-turbo | US | 64K | $0.40 | $1.30 | | novita/deepseek/deepseek-v3.1-terminus | US | 131K | $0.27 | $1.00 | | novita/deepseek/deepseek-v3.2-exp | US | 164K | $0.27 | $0.41 | | novita/deepseek/deepseek-v3/community | US | 64K | $0.89 | $0.89 | | novita/deepseek/deepseek_v3 | US | 64K | $0.89 | $0.89 | | novita/elephant | US | 262K | $0.10 | $0.30 | | novita/glm-4.7 | US | 205K | $0.60 | $2.20 | | novita/glm-4.7-flash | US | 200K | $0.07 | $0.40 | | novita/glm-5 | US | 205K | $1.00 | $3.20 | | novita/google/gemma-3-12b-it | US | 131K | $0.05 | $0.10 | | novita/google/gemma-3-27b-it | US | 98K | $0.12 | $0.20 | | novita/google/gemma-4-26b-a4b-it | US | 262K | $0.13 | $0.40 | | novita/google/gemma-4-31b-it | US | 262K | $0.14 | $0.40 | | novita/gryphe/mythomax-l2-13b | US | 4K | $0.09 | $0.09 | | novita/inclusionai/ling-2.6-1t | US | 262K | $0.30 | $2.50 | USD per 1M tokens. Pricing and availability subject to change. Get started. Paste this into your coding agent (Claude Code, Cursor, Codex, and more) and it will set up Opper and route to Novita for you: Use curl to download, read and follow: https://skills.opper.ai Then set up Opper to use Novita as the provider, e.g. novita/deepseek-v4-flash. Prefer a direct call? Opper is drop-in compatible with the OpenAI, Anthropic, and Google SDKs, so one API key and the model string are all you need: 1import OpenAI from "openai"; 2 3const client = new OpenAI({ 4 apiKey: process.env.OPPER_API_KEY, 5 baseURL: "https://api.opper.ai/v3/compat", 6}); 7 8const completion = await client.chat.completions.create({ 9 model: "novita/deepseek-v4-flash", 10 messages: [{role: "user", content: "Hello"}], 11}); 12console.log(completion.choices[0].message.content); About Novita Novita is an AI cloud platform headquartered in San Francisco, serving 200+ open-weight and multimodal models across text, image, audio, and video through an OpenAI-compatible API, alongside dedicated and serverless GPU infrastructure and agent sandboxes. Novita is SOC 2 Type II audited and does not train on customer data. About Opper AI Opper AI is the European AI gateway and control plane for agents: one EU-hosted, GDPR-compliant API across 300+ models, with smart routing, automatic fallbacks, built-in evaluations and observability, real-time guardrails, and full OpenAI SDK compatibility.

Novita AI has launched Sandbox, a secure infrastructure platform for autonomous AI systems including OpenClaw and Hermes Agent. The platform addresses security risks as AI systems evolve from copilots to autonomous agents that execute code and interact with live environments. Sandbox uses Firecracker microVMs to isolate each agent task with dedicated kernels, isolated memory and ephemeral filesystems, preventing credential leakage and cross-agent interference. The system maintains sub-200 millisecond startup times whilst scaling to thousands of parallel microVMs. The platform supports stateful execution, allowing environments to pause and resume with full runtime state restored in approximately one second. Sandbox is available through managed NovitaClaw deployments, OpenClaw skills and Hermes-compatible runtimes. Novita AI positions the launch as creating a new category: Secure Agent Infrastructure for production-scale autonomous systems.

Novita AI ranked as the best performing & reliable inference layer. PR Newswire Today at 8:07am PDT 120+ LLMs through a single API, with day-0 model availability, OpenAI and Anthropic compatibility, and top-ranked performance validated by Artificial Analysis. SAN FRANCISCO, April 21, 2026 /PRNewswire/ - As demand for open-source AI infrastructure grows, Novita AI is establishing itself as the inference provider for developers and engineering teams that need fast and affordable inference for production AI. The platform covers more than 120 large language models through a single OpenAI-compatible and Anthropic-compatible API, makes every new model available on release day, and ranked #1 for scientific reasoning accuracy across all major inference providers, according to independent benchmarking by Artificial Analysis. Novita AI is trusted by leading teams across the AI ecosystem, including Hugging Face, Quora, OpenRouter, Vercel, Kilo Code, and Genspark. "Open-source AI moves at a pace that most infrastructure hasn't kept up with," said Junyu Huang, COO of Novita AI. "We built Novita to close that gap. When a new model ships, developers can be in production with it the same day, on infrastructure they can actually rely on." Artificial Analysis provides comparison and analysis of AI models and API hosting providers, with independent benchmarks across key performance metrics including quality, price, and output speed. In its GPT-OSS 120B assessment covering all major inference providers, Novita AI ranked as follows (April 2026): GPQA Diamond (scientific reasoning): #1 among all providers, scoring 79.0% across 16 runs AIME 2025 (advanced mathematics): 93.3% across 32 runs, at the level of the top providers IFBench (instruction following): #5, scoring 68.9%, within 0.8 points of the top provider New models ship constantly. Novita AI makes each one available through its API on release day, without exception. For engineering teams running evaluation pipelines or production systems that depend on current models, access is never the bottleneck. Novita AI hosts more than 120 LLMs across every major model family, including Qwen, DeepSeek, LLaMA, Mistral, Gemma, GLM, Phi, and more. All models share the same API format, authentication, and SDK. Teams on the OpenAI or Anthropic SDK can switch to Novita by changing the base URL. Novita's API works out of the box with Claude Code, OpenClaw, Codex CLI, and OpenCode. Novita AI delivers fast inference with the full feature set production AI teams depend on, with no tiered restrictions or add-ons. Tool calling: compliant with OpenAI and Anthropic function-calling specifications, supporting multi-turn agent workflows Structured outputs: JSON responses that conform to a specified schema, no parsing wrappers needed Prompt caching: lower latency and token costs for RAG pipelines and agent sessions with repeated context Novita AI is an AI and agent cloud platform helping developers and startups build, deploy, and scale models and agentic applications with high performance, reliability, and cost efficiency. The platform delivers fast inference across 120+ LLMs and multimodal models through a single API, alongside GPU Instances, Bare Metal, and Agent Sandbox infrastructure built for production AI. SOURCE Novita AI This is a paid placement. For further inquiries, please contact PR Newswire directly.