OpenModel

@lllddd8888 Thanks! Really appreciate the kind words. MiniMax-M3 is a beast for coding and long-context work — super excited to see it getting real-world love. Have you tried it on OpenModel yet? Would love to hear your thoughts!

@openmodel_ Huge milestone for open-source AI, real-world utility and dev experience are finally aligning perfectly.

⚡MiniMax-M3 is now live on OpenModel! - State-of-the-art Coding & Agent performance - 1 Million ultra-long context window - Native image + video understanding All accessible with **one API key** through OpenModel’s production-grade gateway. Smart routing • Automatic failover • Zero downtime Try it now → console.openmodel.ai

At some point the AI stops adapting to you and you start adapting to the AI.

Try Opus 4.8 now: openmodel.ai/model-pricing

Opus 4.8 live on Openmodel! On OpenModel just use the model name: → anthropic/claude-opus-4.8 → Fast Mode Zero code changes. Native tool calling works perfectly with Grok V9, GPTs, and everything else in the same loop. Multi-model just got smoother.

Claude Model Family Reality.jpg

Quick rule of thumb for picking models: - Cheap & fast: DeepSeek V4 Flash ($0.14/1M) - Long context: Claude Opus 4.7 - Code generation: GPT-5.4-codex - Multimodal: Gemini 3 Pro - Tool-heavy agents: Claude Sonnet 4.6 Switch between them with one base URL change. openmodel.ai

@degenpark_eth Yeah, that's the part most gateways quietly punt on. Streaming + tool_use across formats was a fun rabbit hole. Appreciate you testing it in a real loop

@openmodel_ tool schema breakage kills most mixed setups quick. good to see it handled cleanly here

One API. Native OpenAI + Anthropic + Gemini SDKs. Change one base URL → use any model with your existing code. No rewrites. No format conversion. $10 free credits. Sign up by May 30.

Docs: docs.openmodel.ai

One subscription to rule them all is kind of our vibe. You should really check us out if you haven't already!

This is exactly what we hoped people would notice 🙏 The "agent loop with mixed providers" case was the hardest to get right — most gateways break tool schemas between OpenAI ↔ Anthropic ↔ Gemini formats. What's your loop doing, if you can share? Always curious what people build with cross-provider routing.

@openmodel_ switched an agent loop to their base url, gemini handled the same openai tool calls without any format tweaks

@elonmusk @elonmusk This is huge for coding! When Grok V9 drops, OpenModel will let devs route to it by model name with zero code changes — full native tool calling preserved across Grok + Claude + GPT etc. Perfect timing for multi-model workflows.

Grok foundation model V9-Medium (1.5T) has finished training. Evals look good. A lot of Cursor data was added in supplementary training and there is more to come. Fine-tuning is underway and reinforcement learning begins in a few days. 2 to 3 weeks to public release. This will be a major improvement over the 0.5T v8-small that currently serves all Grok production traffic, especially for difficult coding tasks.

@Twin790 Appreciate it! Small correction: not a day pass. Sign up by May 30 → $10 credits, 7-day validity. That's enough for ~thousands of GPT-5 calls or millions of DeepSeek tokens depending on what you test.

@openmodel_ Thanks but it’s best you say a free day pass.

"How does tool calling work when you mix providers?" Best question we got this week. Answer: We preserve the original tool format (OpenAI tools / Anthropic tool_use) when routing. So GPT-5.5 → Claude in the same loop just works. No schema rewrites needed.

@LewsmithUk Fair question 😅 The name reflects our goal of being an "open" gateway to all major LLMs (one API, no provider lock-in), not "open source". Maybe we should clarify this on the site — appreciate the feedback.

@openmodel_ What makes it "open"?

$10 in credits to try OpenModel — multi-model gateway with native OpenAI, Anthropic, and Gemini endpoints. Sign up by May 30. Claude (Haiku/Opus/Sonnet), Gemini 3.x Flash & Pro, GPT-5.3/5.4 series, DeepSeek v4 — all live now.

@johnnybuildr Currently text + vision models only (we cover Claude, GPT, Gemini, DeepSeek vision). Image generation (DALL-E/Imagen/Flux) is on the roadmap. Anything specific you need?

@openmodel_ Does it cover image generation?

Great question. Tool calling works across providers — we preserve the original format (OpenAI tools or Anthropic tool_use) when routing. So you can switch GPT-5 ↔ Claude in the same loop without rewriting your tool schemas. Happy to share more details if you want to test it:docs.openmodel.ai/en

@openmodel_ how does tool calling hold up when you switch between the providers in one loop?

Early access extended. $10 in credits to try OpenModel — multi-model gateway with native OpenAI, Anthropic, and Gemini endpoints. Sign up by May 30. Claude (Haiku/Opus/Sonnet), Gemini 3.x Flash & Pro, GPT-5.3/5.4 series, DeepSeek v4 — all live now. Preview round: credits expire in 7 days, accounts wiped before public launch. → openmodel.ai

Fresh multi-provider model table on OpenModel: • Claude: strong but pricey • Gemini: great value + long context • GPT-5 series: comprehensive • DeepSeek v4: dirt cheap All accessible via your existing SDKs — just change the model name. Native protocol translation + intelligent routing. Zero lock-in. openmodel.ai ($10 free)