Kimi K2.7 Code vs Claude Fable 5: Landing pages that cost 94% less
- Source
- together.ai
- Author
- Together AI
- Date
Context
Together AI's 12-page experiment found a substantial cost difference between Kimi K2.7 Code and Claude Fable 5, alongside lower scores for Kimi. The company reports roughly 94% lower generation costs for Kimi on these landing pages. Quality scores came from GPT-5.5 reviewing screenshots and source code against a rubric, rather than a blind human . The comparison publishes the score table.
Adding screenshots and design references through a tool server had mixed results in that table: Kimi's SQL-chart page rose from 82 to 86, while its architecture portfolio fell from 88 to 79. The experiment supports a cost-quality tradeoff for these generated pages, with visual references helping some outputs and hurting others.
- eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
“We ran 12 landing pages through Kimi K2.7 Code and Claude Fable 5. Kimi cost 94% less and scored within a few points on nearly every page. Open-source models aren't just cheaper, they're genuinely competitive on quality. And the gap is closing faster than people realize.”
Together AI
“For example, this landing page for a B2B SaaS cost just 4 cents with Kimi. The same prompt cost $1.09 with Claude Fable, making it almost 27 times more expensive.”
Together AI
“On average Kimi was ~16x cheaper than Fable and ~8x cheaper than Opus.”
Together AI
“Unfortunately, both models made landing pages that felt recognizably AI-generated.”
Together AI
“The biggest improvement came from giving Kimi visual inspiration through a custom MCP server. Once it could work from screenshots and design references, the pages became more readable, more structured, and more visually intentional.”
Together AI
Checking sign-in…
Loading comments…