Two Chinese-lab models that compete on value. DeepSeek V4-Pro is the cheaper API by a wide margin; Kimi K3 is a higher-tier open-weights agentic model (now Moonshot's previous-generation flagship) with constant pricing.
DeepSeek V4-Pro is about 4.5× cheaper on input and 7.5× on output than Kimi K3 off-peak, and still cheaper at peak. Kimi K3 sits a tier higher (Moonshot's previous-generation flagship, behind K3.1 Pro), has open weights, and doesn't vary its price by time of day.
| DeepSeek V4-Pro | Kimi K3 | |
|---|---|---|
| Provider | DeepSeek | Moonshot AI |
| Tier | Mid-tier | Flagship |
| Input / 1M tokens | $0.66 | $3.00 |
| Output / 1M tokens | $1.98 | $15.00 |
| Cache hit / 1M | $0.022 | $0.30 |
| Cache write / 1M | — | $3.00 |
| Context window | 1M | 1M |
| Notes | A planned Sep 14, 2026 reroute to V4.1-Flash rates was reversed after user pushback; billing unchanged since the Aug 13, 2026 GA build. Off-peak rate shown; peak (01:00-04:00 & 06:00-10:00 UTC weekdays) is 2x: $1.32/$3.96 | Previous-generation flagship, superseded by Kimi K3.1 Pro but still sold; open weights; cache write $3 (5-min TTL) or $6 (1-h TTL) |
DeepSeek V4-Pro is 4.5× cheaper on input tokens ($0.66 vs $3.00 per 1M). DeepSeek V4-Pro is 7.6× cheaper on output tokens ($1.98 vs $15.00 per 1M).
| Workload | DeepSeek V4-Pro | Kimi K3 | Difference |
|---|---|---|---|
| Support chatbot 10,000 conversations × (2,000 input + 400 output tokens) | $21.12 | $120 | DeepSeek V4-Pro saves $98.88 |
| Document Q&A 1,000 questions × (30,000 input + 800 output tokens) | $21.38 | $102 | DeepSeek V4-Pro saves $80.62 |
| Coding agent session 4M input tokens (90% cache hits) + 150,000 output tokens | $0.64 | $4.53 | DeepSeek V4-Pro saves $3.89 |
Standard rates per 1M tokens from data on this site. The coding-agent row assumes 90% of input is read from the prompt cache (full input price where a model has no cache rate). Cache-write fees, batch discounts and long-context surcharges listed in the notes are not included. Updated November 2, 2026.
you want the lowest API price for text work and can shift bulk jobs outside DeepSeek's weekday peak window.
you want a higher-tier agentic model with open weights, a constant hourly price, and batch inference at 60% of standard.
DeepSeek's discount lever is time: rates double from 01:00–04:00 and 06:00–10:00 UTC on weekdays, so off-peak scheduling is the cheap path. Moonshot's lever is batch: kimi-k3 batch inference is 60% of standard price. Even batched, Kimi K3 costs more than off-peak V4-Pro.
V4-Pro's cache hit is $0.022 off-peak; Kimi K3's is $0.30, and Moonshot also bills cache writes separately by TTL (5 minutes or 1 hour). For agents that resend a long context, V4-Pro's cached tokens are an order of magnitude cheaper.
Kimi K3 is a former flagship (Moonshot now leads with K3.1 Pro) and V4-Pro is mid-tier, so run your own evals before assuming they're interchangeable. Kimi K3's open weights let you move it off Moonshot's API later. Both offer about 1M tokens of context; DeepSeek lists V4-Pro as text-only with up to 384K output tokens.
DeepSeek V4-Pro, at both off-peak and peak rates. Kimi K3's batch rate (60% of standard) still doesn't reach V4-Pro's off-peak price.
No time-of-day pricing is listed for Kimi K3. DeepSeek V4-Pro doubles its rates during weekday peak hours.
Yes, Kimi K3 is an open-weights model.
How we choose: prices come from each vendor's official pricing page and are re-checked every week. We take no affiliate fees, so rankings aren't for sale.