API Pricing Comparison

Per-million-token costs for China's leading open-source coding models — where LongCat-2.0's pricing is one of its strongest differentiators today.

Standard API Pricing (USD per 1M tokens)

Model Input ($/M) Output ($/M) Cached Context Notes
LongCat-2.0 (standard) 0.75 2.95 Free on cache hit 1M context; agentic coding focus
LongCat-2.0 (promo) 0.30 1.20 Free on cache hit Launch promotion pricing
GLM-5.1 ~1.40+ Varies by tier Check provider Frontend / web-agent scenarios from ~$1.40/M input
Kimi K2.6 ~0.95+ Varies by tier Check provider 256K context; long-session agent workloads
DeepSeek V4-Pro Check provider Check provider Check provider Pricing varies by API gateway; compare at OpenRouter
Qwen3.6 Plus Check provider Check provider Check provider Often positioned for cost-efficient terminal tasks

Prices reflect publicly reported standard and promotional rates. Always verify current pricing on longcat.ai, OpenRouter, or your API provider before budgeting.

Example Workload Costs

Estimated cost for a typical agentic coding session: 100K input + 20K output tokens (e.g., multi-file refactor with tool calls).

Model Standard Rate Promo Rate (LongCat only)
LongCat-2.0 $0.075 + $0.059 = ~$0.13 $0.030 + $0.024 = ~$0.05
GLM-5.1 (input ~$1.40/M) ~$0.14+ input alone
Kimi K2.6 (input ~$0.95/M) ~$0.095+ input alone

For high-volume agent pipelines with repeated context (system prompts, codebase snapshots), LongCat's free cached-context reads can reduce effective input cost significantly — a factor flat per-token tables often miss.

When Price Matters Most

  • Agent loops: Multi-turn tool calling inflates token counts — output pricing dominates
  • Long-context RAG: 1M context models send large prompts; cache hits save real money
  • Production scale: At millions of tokens/day, $0.30 vs $0.95/M input is a 3× difference
  • Evaluation & fine-tuning: Cheap inference enables more iteration on prompts and agents

Price is not everything — benchmark fit matters too. See our positioning guide and head-to-head pages for performance trade-offs.