Key Takeaways

• Tencent’s Tangent is a 295B parameter model released free with commercial Apache 2.0 licensing — no restrictions
• Scored 78 on coding benchmarks, beating many closed-source competitors
• Hallucination rate dropped from 12% to 5% — significantly more honest than typical models
• Already deployed in real products: WeChat assistants, Yuanbao, and video games
• Free access through OpenRouter until July 1st (later extended to July 21st)
• This is part of a deliberate Chinese strategy: free models, cheap infrastructure, ecosystem lock-in

What Tangent Actually Delivers: The Numbers

When I first pulled up the benchmarks, I had to double-check I was reading them right. Tangent scores 78 on coding tests — that’s above many closed models that charge premium prices. On science and reasoning benchmarks, it hits over 90 points. But the figure that genuinely impressed me was the hallucination rate.

Tencent claims they reduced the “made-up answer” rate from 12% down to 5%. In my experience testing dozens of models, hallucination is the silent killer of productivity — you don’t notice it until a wrong answer costs you time or credibility. A 5% rate puts Tangent in genuinely trustworthy territory for research and coding workflows.

The model also jumped from 35% to 45.6% on certain reasoning benchmarks, surpassing DeepSeek in specific categories. Looking at the comparison charts across LRM (Large Reasoning Model) evaluations, Tangent sits at or near the top alongside Qwen and Cloud models. What struck me was seeing a completely free, open-weight model ranking first in some categories against paid competitors like GPT-4, Claude Opus, Gemini, and others.

Why “Free” Here Actually Means Free

I’ve learned to be skeptical of “free” AI announcements. Usually there’s a catch — rate limits, non-commercial clauses, or the model quietly disappears. Tangent’s license is Apache 2.0. That means you can download it, modify it, run it commercially, embed it in products — with zero restrictions. No attribution requirements that cripple usage, no enterprise licensing traps.

This isn’t theoretical. The model is already running in WeChat’s own AI assistants, in Yuanbao, and even inside a video game. This is real production deployment, not a research demo. When I see models benchmarked on LLM Arena or other leaderboards, I always check whether they’ve been tested on real users or just synthetic tests. Tangent has actual user volume behind it.

How I Accessed Tangent (No Downloads Required)

Here’s the practical part I know most readers want. You don’t need to download 295 billion parameters to your laptop. I accessed Tangent entirely through OpenRouter, and it was genuinely free when I tested it.

The process: search for “Tangent” on OpenRouter, look for the model marked “free” (not the paid tier running at roughly $0.20-$0.80 per million tokens). Create an API key, then integrate it into whatever interface you prefer — I used it through Claude Code, but it works with OpenCat, various agent frameworks, or direct API calls. OpenRouter’s interface breaks down which models excel in which domains, which browsers support them, and real usage statistics.

Important caveat: When I tested, the free tier was available until July 1st. Based on the pattern with similar releases, I expect this could shift to paid tiers eventually — though the transcript later confirmed extension to July 21st. If you’re reading this after that date, check OpenRouter directly for current pricing. The paid version, even if it becomes necessary, still runs dramatically cheaper than Western alternatives.

The Honest Reality: Is It “The Best” Model?

I want to be straight with you here, because I see other coverage claiming Tangent dethrones everything. In my actual usage and based on the deeper evaluations I trust, Claude Opus remains the strongest general-purpose model for most complex tasks. On LLM Arena’s value-per-token analysis, Claude still commands premium pricing because it delivers premium results.

Where Tangent changes the game is the price-to-performance ratio. When a model scoring 78 on coding and 90+ on reasoning costs literally nothing, while competitors charge $7+ per million tokens, the economics become absurd. For prototyping, automation workflows, agent systems, and any application where you need scale rather than absolute peak quality, Tangent is now my default recommendation.

I run mixed setups in my own work — Claude for critical reasoning tasks, Tangent and similar models for volume operations, specific fine-tuned models for niche domains. The strategy isn’t finding one “best” model anymore. It’s building a portfolio based on cost, capability, and risk tolerance.

The Bigger Pattern: Why This Keeps Happening

Here’s what I find more significant than any single benchmark. Yesterday: food delivery company releases free AI. Today: Tencent releases free AI. These aren’t coincidences — they’re coordinated strategic moves.

Consider the contrast. America’s strongest models — GPT-4, Claude, Gemini — are closed, expensive, and politically vulnerable. Remember what happened with Fabri 5? Shut down, reopened, facing potential closure again with days of notice. Your infrastructure dependency can disappear overnight based on regulatory whim.

China’s strategy is inverted: free models, cheap infrastructure, ecosystem habituation. Release powerful weights at no cost. Offer inference cheaper than anyone else. Get global developers building on your stack. Once the dependency exists, the economic relationship follows.

Food delivery, gaming, pure technology companies — different sectors, same playbook. Train the model, prove it beats paid alternatives on benchmarks, then price it at disruption levels. I expect this pattern to accelerate through the coming months. Each announcement will look like separate “new model” news. The reality is a coordinated competitive war where developers and small businesses are the intended beneficiaries — at least in the short term.

What This Means for Your Actual Work

The practical implication I keep coming back to: powerful AI tools are entering your pocket for free. Not subsidized-free. Not free-trial-free. Actually free, with commercial rights, with proven deployment at scale.

For my own automation projects — eBay listing generation, Facebook ad copy, Etsy product descriptions, Amazon review analysis, content systems — I’ve been systematically testing where Tangent-level quality is sufficient. Often, it is. The cost savings let me run more experiments, test more variants, operate at scales that would be economically impossible with premium API pricing.

The caveat I always include: free tools don’t guarantee results. Your prompt engineering, workflow design, quality validation, and business strategy still determine outcomes. But the barrier to experimentation has never been lower.

FAQ

Is Tangent really completely free for commercial use?

Yes. Tencent released Tangent under the Apache 2.0 license, which permits unlimited commercial use, modification, and distribution without restrictions. The free tier on OpenRouter was available through July 21st; check current availability as this may shift to paid tiers.

How does Tangent compare to GPT-4 and Claude?

On specific benchmarks, Tangent surpasses many paid models — 78 on coding tests, over 90 on science/reasoning, and reduced hallucination to 5%. However, Claude Opus still leads for complex general reasoning in my testing. Tangent’s advantage is delivering near-top-tier performance at zero or minimal cost.

Do I need technical skills to use Tangent?

Basic technical comfort helps. The simplest path is OpenRouter’s web interface with an API key. For integration into workflows, you’ll need familiarity with API calls or tools like Claude Code, OpenCat, or agent frameworks. No model download or local setup required.

Why are Chinese companies releasing free AI models?

This appears to be strategic ecosystem building: attract global developers with free, capable tools, establish infrastructure dependencies, and compete with closed American models on price. The companies involved span food delivery, gaming, and pure technology — suggesting coordinated national-level competitive positioning rather than isolated generosity.

Conclusion

I tested Tencent’s Tangent because the numbers demanded attention — 295 billion parameters, 78 coding score, 5% hallucination rate, zero cost. What I found was a genuinely capable tool that fits into a growing portfolio of free, powerful AI options. It’s not magic, and it’s not universally superior to premium alternatives. But it represents something important: the continued demolition of the assumption that better AI must cost more.

The trend is clear. The strategic pattern is clear. For builders, automators, and small business owners willing to test and validate, the tools available at zero cost today would have cost thousands in API fees just months ago. My recommendation: test Tangent while the free access lasts, build your evaluation framework, and prepare for more announcements following this same pattern. The AI infrastructure landscape is being rewritten in real time — and for once, the immediate beneficiaries look like individual developers rather than platform owners.

Watch the full video (in Turkish — English subtitles available):

Tools & Community

• TurkoLister — the AI listing tool I use to turn Amazon products into optimized eBay UK listings in about 60 seconds (from £4.99/month, £1 one-week trial).
• AI & E-commerce Community — my Turkish-speaking community ($19/month) with weekly live sessions.
• Subscribe on YouTube — new experiments every week.

The post Tencent’s Free 295B AI Model Beat GPT-4 — I Tested It appeared first on Digital Market Mentoring.

Yesterday, the mask dropped. The creator wasn’t a Silicon Valley lab. It was Meituan—yes, the Chinese food delivery giant. Think of it as their equivalent of Yemeksepeti or Deliveroo. Through their AI division Kyutai, they had secretly trained one of the world’s largest open-source models: 1.5 trillion parameters, 1 million token context window, trained entirely on Chinese Cambricon chips—not a single Nvidia GPU involved.

This matters for anyone building with AI, and I’ll explain exactly why—including how to use it for pennies.

Key Takeaways

• 1.5 trillion parameters with 1 million token context—quality near Claude Sonnet at roughly 1/10th the price
• Trained on 50,000 Cambricon cards, making it the first trillion-parameter model built without American hardware
• MIT licensed—fully open source, free for commercial use
• Input tokens cost roughly $0.30 per million versus Claude’s ~$3.00
• Free context caching: re-read the same project files without paying again
• Available via OpenRouter, Kyutai’s own API, or through Kimi’s Client multi-agent system
• Weights not yet downloadable—API-only for now, with local deployment coming

The Mystery Model That Took Over OpenRouter

I noticed On Alpha the same way most developers did: it just appeared. No announcement, no blog post, no corporate branding. On OpenRouter—a platform I use regularly to compare models—it started climbing the rankings with disturbing speed.

Here’s what caught my attention:

• Code generation quality that matched or exceeded top-tier models
• Pricing that seemed like a bug—not a feature
• 1 million token context window, meaning it could ingest entire codebases, hundreds of pages of documentation, or massive project files in one go

Usage jumped 242% as word spread through developer channels. Meanwhile, Claude—previously dominant on the platform—slipped to #2. People were unknowingly adopting a Chinese model, and nobody knew who to thank (or blame).

The reveal came when Moonshot AI (the company behind Kimi) publicly confirmed what investigators had suspected: On Alpha was built by Kyutai, Meituan’s AI research lab. A food delivery company had created one of the most capable open models in existence.

Why the Hardware Story Changes Everything

The technical achievement here goes deeper than model architecture. Kyutai trained this 1.5 trillion parameter system on 50,000 Cambricon AI chips—Chinese-designed, Chinese-manufactured processors. No Nvidia H100s. No A100s. No American hardware at any stage.

This is historically significant. For years, US export controls on advanced semiconductors were treated as a hard ceiling on Chinese AI development. When Fablo 5 (apparently a reference to a previous model or service) was restricted, American users lost access for three weeks. The assumption was that without Nvidia chips, you couldn’t compete at the frontier.

Kyutai just proved that assumption wrong. A food delivery platform’s research team built a trillion-parameter model using domestic alternatives. The rules of this game are rewriting themselves in real-time.

For entrepreneurs like me, based in the UK but working globally, this means hardware diversification is accelerating. More training pipelines, more model providers, more resilience against single-point-of-failure restrictions. The monopoly on cutting-edge AI is cracking.

Real Numbers: What This Costs vs. Claude

Here’s where this becomes immediately practical. I run multiple automation workflows and AI-assisted development projects. Token costs are a real line item in my monthly expenses.

That’s roughly a 10x price difference for comparable quality. When you’re running thousands of API calls monthly—processing documents, generating code, analyzing data—this isn’t marginal savings. It’s the difference between a $300/month AI bill and a $30 one.

Kyutai’s current promotion drops their already-low pricing even further. Their token packages (prepaid, 30-day validity) beat pay-as-you-go API pricing for consistent usage. During launch, they’re offering what was normally a $299 tier for $60.

But the feature that genuinely impressed me: free context caching. When you’re working on the same project repeatedly—refining code, iterating on documents—the model doesn’t charge you again for re-reading the same context. This is how it should work, and frankly, Western providers should take note. For long-running projects with extensive context, this alone can cut costs by 30-50%.

How I’m Actually Using It: Three Pathways

Option 1: OpenRouter (Simplest)

If you already have an OpenRouter account, search for the model directly and select it. No additional setup. This is how I first tested it—took under two minutes to start making calls.

Option 2: Kyutai’s Native API

For direct access, create an account on Kyutai’s platform, generate an API key, and integrate. They’ve made this deliberately compatible with OpenAI and Claude SDK formats, so if you’re already using those clients, you typically just change the base URL. I’ve tested this in Cursor and standard API clients—it works cleanly.

Option 3: Kimi Client (Most Powerful for Multi-Agent Work)

This is where it gets interesting for serious builders. Kimi’s Client (from Moonshot AI) has evolved into something I haven’t seen elsewhere: a multi-agent routing system that automatically selects the best Chinese AI model for each specific task.

Here’s how it works in practice: you submit a project, and Client dynamically routes to DeepSeek, MiniMax, Kimi itself, or now Kyutai’s models—whichever performs best for that particular job. I’ve watched it switch between models mid-workflow based on task characteristics.

Setup is straightforward:

• Install the Kimi Client extension in VS Code (most popular option)
• Or run directly from terminal using their provided code snippets
• Enable the extension, and free-tier models auto-select without configuration

For my automation workflows, this eliminates the manual model-selection overhead I used to spend significant time on.

Current Limitations I Need to Flag

I’m not going to oversell this. There are genuine constraints:

• Weights unavailable for download—API-only currently. Kyutai says local weights are coming; I’ll test when they arrive.
• Quality ceiling: In my testing, it doesn’t consistently beat Claude 3.5 Opus (Anthropic’s absolute strongest model). It’s competitive with Sonnet-level performance—excellent for daily work, but not universally superior.
• Language quirks: The voice and mobile interfaces I tested defaulted to Chinese. English text interfaces work fine, but don’t expect seamless multilingual voice interaction yet.
• Documentation is Chinese-first—browser translation handles this, but it’s friction.

For the price, these are manageable trade-offs. But manage your expectations: this is a Claude Sonnet competitor at Claude Haiku pricing, not a free lunch.

The Bigger Picture for Builders

What Meituan/Kyutai achieved here signals something I predicted in my community discussions: AI capability is decentralizing faster than consensus expects. A food delivery company trained a frontier-class model. They did it without the hardware everyone assumed was mandatory. They released it under MIT license, letting anyone build commercial products on top.

For my own projects—automating e-commerce operations, building AI-assisted content workflows, training specialized agents—this expands the viable toolset dramatically. When I mentor developers in our live sessions, I emphasize cost sustainability: your AI infrastructure needs to survive your revenue ramp-up period. Models like this make that math work.

The companies actually building with AI right now—not just consuming chat interfaces—are going to benefit most from this wave. The moat isn’t access to expensive models anymore. It’s knowing how to compose, route, and deploy the right models for specific outcomes.

FAQ

Is Kyutai On Alpha really free to use commercially?

Yes. It’s released under an MIT license, which permits commercial use, modification, and distribution without restriction. When weights become available for download, you’ll be able to self-host for internal products. Currently, API usage has standard metering but no licensing fees.

How does the quality compare to Claude 3.5 Sonnet?

In my testing, it’s comparable to Sonnet-level performance for coding and analysis tasks, but doesn’t consistently exceed Claude 3.5 Opus (Anthropic’s top tier). The 1 million token context window matches or exceeds most Claude tiers. For daily development work, document analysis, and automation—it’s genuinely competitive. For frontier research or the most demanding reasoning tasks, Claude Opus still holds an edge.

Can I use this if I only speak English?

Text interfaces and API calls work fully in English—that’s how I’ve been using it. The web dashboard and documentation are Chinese-first, but browser translation handles this adequately. Voice features and some mobile app functions currently default to Chinese. If you’re comfortable with API integration or using OpenRouter as an intermediary, language isn’t a blocker.

What’s the cheapest way to get started?

OpenRouter with existing credits is fastest—no new account needed. For dedicated usage, Kyutai’s token packages (prepaid, ~$60 promotional tier) beat pay-as-you-go API pricing for consistent workloads. Their free context caching means repeated work on the same project costs progressively less. If you’re experimenting, start with OpenRouter; if you’re building a production workflow, the direct API with token packages optimizes costs.

Conclusion

I tested this model because the numbers didn’t make sense—a ghost topping leaderboards at impossibly low prices. Two months of quiet usage by developers like me validated the quality before the corporate reveal.

What Meituan’s Kyutai built isn’t just a cheaper Claude alternative. It’s proof that AI model training is escaping its hardware cage, that open-source licensing is becoming a competitive weapon, and that the companies willing to operate in secrecy for months can reshape market dynamics overnight.

For my daily work, this joins my toolkit alongside Claude, GPT-4, and specialized open models. The routing logic matters now more than model loyalty. I use what delivers the right quality at sustainable cost for each specific task.

The weights aren’t downloadable yet. When they are, I’ll run local benchmarks and share results. Until then, the API is production-ready, the pricing is genuinely disruptive, and the context window handles projects that would chunk and degrade on smaller models.

If you’re building AI-powered systems and haven’t pressure-tested Chinese models recently, you’re working with an incomplete picture of what’s possible. The food delivery company just schooled the pure-play AI labs on cost-engineering at scale. That’s worth paying attention to.

Watch the full video (in Turkish — English subtitles available):

Tools & Community

• TurkoLister — the AI listing tool I use to turn Amazon products into optimized eBay UK listings in about 60 seconds (from £4.99/month, £1 one-week trial).
• AI & E-commerce Community — my Turkish-speaking community ($19/month) with weekly live sessions.
• Subscribe on YouTube — new experiments every week.

The post How a Food Delivery Giant Built a $0.30 AI Model to Rival Claude appeared first on Digital Market Mentoring.