Kimi K3 vs Claude Opus 4.8: Benchmarks, Price, Verdict

Kimi K3 is the first open-weight model that makes the Kimi K3 vs Claude Opus 4.8 question worth asking seriously. On the Artificial Analysis Intelligence Index, K3 ranks fourth of 189 models — level with Opus 4.8 and GPT-5.5, behind only Claude Fable 5 and GPT-5.6 Sol. It also costs 40% less per token, and its weights are expected to be released by July 27.

The honest answer is not "K3 wins" or "Opus wins" — it depends on what you're optimizing for. Here are the numbers, then the verdict.

Benchmarks: Kimi K3 vs Claude Opus 4.8

*Reported on different harnesses (Opus 4.8's score uses the Terminus-2 harness), so treat the Terminal-Bench gap as directional, not exact.

Three takeaways. On graduate-level reasoning (GPQA Diamond) the models are statistically tied. On agentic terminal work, K3's published number is well ahead, with the harness caveat above. And in Arena's blind Frontend Code testing, developers ranked K3 first outright — ahead of every model, including Anthropic's.

Where Opus 4.8 keeps an edge: SWE-bench Verified at 88.6% is a published, battle-tested result K3 has no counterpart for yet, and Anthropic's models remain the default target for agent harnesses — Claude Code, MCP tooling, and most agentic scaffolds are tuned against Claude first. K3 is three days old; its ecosystem is not.

Pricing: 40% cheaper across the board

Per-token rates through LLM Gateway (each provider's published pricing):

Both have a 1M-token context window. K3's output limit defaults to 131K tokens but is configurable up to the full 1M in a single response; Opus 4.8 caps output at 128K.

Concrete math: a coding-agent workload of 100M input and 20M output tokens a month runs $600 on Kimi K3 and $1,000 on Claude Opus 4.8. Same ratio at any scale — K3 is 40% cheaper on input, cached input, and output alike.

One operational difference: K3's reasoning is always on at full effort — reasoning_effort currently accepts only max, with lower-effort modes promised in later updates. That is part of why its benchmark numbers are strong, but it means K3 spends thinking tokens even on trivial requests. Opus 4.8 lets you dial reasoning effort up and down per request today — cheaper and faster on the easy 80%.

What each one is for

Pick Kimi K3 when cost dominates, you want frontend-heavy code generation (the Arena result is real), you need single-shot outputs longer than 128K tokens, or open weights matter — self-hosting, fine-tuning, or simply not depending on one vendor's API terms.

Pick Claude Opus 4.8 when you're running mature agent harnesses tuned for Claude, you want configurable reasoning effort, or you need the SWE-bench-class repo-editing reliability that teams have already validated in production.

Or don't pick. The switch between them is a one-word change to the request body, which makes the real answer an A/B test on your own workload:

curl https://api.llmgateway.io/v1/chat/completions \

  -H "Authorization: Bearer $LLM_GATEWAY_API_KEY" \

  -H "Content-Type: application/json" \

  -d '{

    "model": "kimi-k3",

    "messages": [{"role": "user", "content": "Implement the retry logic described in RFC-42.md"}]

  }'

Run a week on kimi-k3, a week on claude-opus-4-8, and read the answer off your LLM Gateway dashboard — per-model cost, latency, and token counts side by side.

Flat rate or pay-as-you-go

Both models count as premium-tier on for the full release details.

Getting started

• 
  — both models on one flat rate, from $29/mo


• Wire K3 into your editor with .
  Vollständiger Original-Bericht

  Ausführliche Details, Code-Beispiele & Hersteller-Stellungnahme auf dev.to.

  ↗ Original-Artikel auf dev.to lesen