Model comparison

DeepSeek vs Kimi API price, context and cost.

DeepSeek (DeepSeek) against Kimi (Moonshot) on RouterPlex, flagship to flagship and cheap model to cheap model: DeepSeek V4 Pro vs Kimi K3.1 Pro, then DeepSeek V4.1 Flash vs Kimi K2.6. A coding-agent session costs $0.90 on DeepSeek V4 Pro and $2.10 on Kimi K3.1 Pro; on the cheap models, $0.25 against $0.91.

Catalog rates17 November 2026
DeepSeek V4 Pro$0.80 / $2.39
Kimi K3.1 Pro$2.00 / $10.00
DeepSeek V4.1 Flash$0.18 / $0.73
Kimi K2.6$0.95 / $4.00
Input / output per 1M tokens

Both lineups

Every live DeepSeek and Kimi model.

Picked from the live catalog by one rule: the flagship is the newest full-size version (no Flash or Lite in the id), the cheap pick is the lowest cost on a 100K-input, 10K-output turn. Speed tiers, abliterated builds and routes without a working upstream are left out. Cheapest first within each family.

ModelRoleContextInput / 1MOutput / 1MImage input
DeepSeek / DeepSeek
DeepSeek V4.1 Flash

deepseek-v4.1-flash

Cheap pick1M$0.18$0.73Yes
DeepSeek V4.1 Flash 1017

deepseek-v4.1-flash-1017

1M$0.18$0.73Yes
DeepSeek V4 Pro

deepseek-v4-pro

Flagship1M$0.80$2.39No
DeepSeek V4 Pro 0813

deepseek-v4-pro-0813

1M$0.80$2.39No
Kimi / Moonshot
Kimi K2.6

kimi-k2.6

Cheap pick256K$0.95$4.00Yes
Kimi K2.7 Code

kimi-k2.7-code

262K$0.95$4.00Yes
Kimi K3.1 Pro

kimi-k3.1-pro

Flagship1M$2.00$10.00Yes
Kimi K3

kimi-k3

1M$3.00$15.00Yes

Flagship tier

DeepSeek V4 Pro vs Kimi K3.1 Pro

Catalog fieldDeepSeek V4 ProKimi K3.1 Pro
ProviderDeepSeekMoonshot
Model IDdeepseek-v4-prokimi-k3.1-pro
Input / 1M$0.80$2.00
Output / 1M$2.39$10.00
Cached input / 1M$0.21$0.19
Context window1M (1,000,000 tokens)1M (1,048,576 tokens)
Max output384K tokens131K tokens
Image inputNoYes
Tool callingYesYes
ReasoningYesYes
Long-prompt tierNoneNone
Price basisVendor list price, 0% markupVendor list price, 0% markup
  • DeepSeek V4 Pro: DeepSeek prices this model by time of day: $0.66 input / $1.98 output per 1M tokens off-peak, and double that during peak hours (01:00-04:00 and 06:00-10:00 UTC, Monday to Friday). RouterPlex bills a single flat rate of $0.80 / $2.39 in between -- the time-weighted average of the two -- so you are never billed by the clock. Cached prompt tokens are billed at $0.21 per 1M, set from what our upstream charges us rather than DeepSeek's cache-hit rate.

What the catalog says

DeepSeek V4 Pro is 60% cheaper on input ($0.80 against $2.00 per 1M tokens). DeepSeek V4 Pro is 76% cheaper on output ($2.39 against $10.00 per 1M tokens). Kimi K3.1 Pro is 10% cheaper on cached input ($0.19 against $0.21 per 1M tokens). On the coding-agent session below, DeepSeek V4 Pro costs $0.90 against $2.10 for Kimi K3.1 Pro. Both accept a 1M context window. DeepSeek V4 Pro can return up to 384K tokens per response, against 131K for Kimi K3.1 Pro. Only Kimi K3.1 Pro lists image input in the catalog; DeepSeek V4 Pro does not.

Generated from catalog prices, limits and capability flags. Price does not measure output quality.

Published benchmarks

Where real data exists.

Raw score as the source publishes it, percentile in brackets. Only rows with a cited source are shown; "Not listed" means the source has no entry for that model.

Source / categoryDeepSeek V4 ProKimi K3.1 Pro
overallLMArena Text Leaderboard1457.8 (87)Not listed
codingLMArena Text Leaderboard1501.8 (84)Not listed
reasoningLMArena Text Leaderboard1445.1 (82)Not listed
writingLMArena Text Leaderboard1444.5 (89)Not listed
generalLMArena Text Leaderboard1453.2 (88)Not listed

LMArena Text Leaderboard, version 2026-09-02, retrieved 2026-09-10. A leaderboard measures its own prompt set, not your workload.

Cheap tier

DeepSeek V4.1 Flash vs Kimi K2.6

Catalog fieldDeepSeek V4.1 FlashKimi K2.6
ProviderDeepSeekMoonshot
Model IDdeepseek-v4.1-flashkimi-k2.6
Input / 1M$0.18$0.95
Output / 1M$0.73$4.00
Cached input / 1M$0.065$0.08
Context window1M (1,000,000 tokens)256K (256,000 tokens)
Max output384K tokens262K tokens
Image inputYesYes
Tool callingYesYes
ReasoningYesYes
Long-prompt tierNoneNone
Price basisVendor list price, 0% markupVendor list price, 0% markup
  • DeepSeek V4.1 Flash: DeepSeek prices this model by time of day: $0.15 input / $0.60 output per 1M tokens off-peak, and double that during peak hours (01:00-04:00 and 06:00-10:00 UTC, Monday to Friday). RouterPlex bills a single flat rate in between -- the time-weighted average of the two -- so you are never billed by the clock and never pay the peak rate for off-peak work.
  • Kimi K2.6: RouterPlex bills $0.95 input / $4 output per 1M tokens, Moonshot's list price. Cached prompt tokens are billed at $0.08 per 1M (Moonshot lists $0.16): our cache price is set from what our upstream charges us, not from the lab's rate.

What the catalog says

DeepSeek V4.1 Flash is 81% cheaper on input ($0.18 against $0.95 per 1M tokens). DeepSeek V4.1 Flash is 82% cheaper on output ($0.73 against $4.00 per 1M tokens). DeepSeek V4.1 Flash is 19% cheaper on cached input ($0.065 against $0.08 per 1M tokens). On the coding-agent session below, DeepSeek V4.1 Flash costs $0.25 against $0.91 for Kimi K2.6. DeepSeek V4.1 Flash has the larger context window: 1M against 256K. DeepSeek V4.1 Flash can return up to 384K tokens per response, against 262K for Kimi K2.6.

Generated from catalog prices, limits and capability flags. Price does not measure output quality.

Published benchmarks

Where real data exists.

Raw score as the source publishes it, percentile in brackets. Only rows with a cited source are shown; "Not listed" means the source has no entry for that model.

Source / categoryDeepSeek V4.1 FlashKimi K2.6
overallLMArena Text LeaderboardNot listed1460.7 (88)
codingLMArena Text LeaderboardNot listed1514.3 (91)
reasoningLMArena Text LeaderboardNot listed1478.4 (94)
writingLMArena Text LeaderboardNot listed1430.8 (85)
generalLMArena Text LeaderboardNot listed1453.6 (88)

LMArena Text Leaderboard, version 2026-09-02, retrieved 2026-09-10. A leaderboard measures its own prompt set, not your workload.

Worked cost examples

Same workload, both bills.

Each line multiplies the catalog rate per 1M tokens by the tokens shown. Cached tokens use the route's cached-input rate, or the full input rate when it has no cache discount. Reasoning tokens bill as output, so a model that thinks at length raises the output line beyond these figures.

Coding-agent session

40 requests of 50K prompt tokens, 80% of each prompt a cache hit, 2.5K output each.

DeepSeek V4 Pro$0.90

0.4M input x $0.80 + 1.6M cached x $0.21 + 0.1M output x $2.39 = $0.90

Kimi K3.1 Pro$2.10

0.4M input x $2.00 + 1.6M cached x $0.19 + 0.1M output x $10.00 = $2.10

DeepSeek V4.1 Flash$0.25

0.4M input x $0.18 + 1.6M cached x $0.065 + 0.1M output x $0.73 = $0.25

Kimi K2.6$0.91

0.4M input x $0.95 + 1.6M cached x $0.08 + 0.1M output x $4.00 = $0.91

Chat app, one month

10,000 conversations of 2K prompt tokens and 500 output tokens, no caching.

DeepSeek V4 Pro$27.95

20M input x $0.80 + 5M output x $2.39 = $27.95

Kimi K3.1 Pro$90.00

20M input x $2.00 + 5M output x $10.00 = $90.00

DeepSeek V4.1 Flash$7.25

20M input x $0.18 + 5M output x $0.73 = $7.25

Kimi K2.6$39.00

20M input x $0.95 + 5M output x $4.00 = $39.00

Long-document Q&A batch

200 requests of 150K prompt tokens and 2K output tokens, no caching.

DeepSeek V4 Pro$24.96

30M input x $0.80 + 0.4M output x $2.39 = $24.96

Kimi K3.1 Pro$64.00

30M input x $2.00 + 0.4M output x $10.00 = $64.00

DeepSeek V4.1 Flash$5.69

30M input x $0.18 + 0.4M output x $0.73 = $5.69

Kimi K2.6$30.10

30M input x $0.95 + 0.4M output x $4.00 = $30.10

Switching

One key, one line.

Both models sit behind the same RouterPlex key and the same OpenAI-compatible endpoint. Keep the request shape and change the model string. Claude Code and the Anthropic SDK use the same ids on https://api.routerplex.com.

Read the quickstart
OpenAI SDK (Python)
import os
from openai import OpenAI
 
client = OpenAI(
base_url="https://api.routerplex.com/v1",
api_key=os.environ["ROUTERPLEX_API_KEY"],
)
reply = client.chat.completions.create(
model="deepseek-v4-pro", # switch: "kimi-k3.1-pro"
messages=[{"role": "user", "content": "Hello!"}],
)
print(reply.choices[0].message.content)
curl /chat/completions
# DeepSeek V4 Pro
curl https://api.routerplex.com/v1/chat/completions \
-H "Authorization: Bearer $ROUTERPLEX_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "deepseek-v4-pro", "messages": [{"role": "user", "content": "Hello!"}]}'
 
# Kimi K3.1 Pro: same request, one field changed
curl https://api.routerplex.com/v1/chat/completions \
-H "Authorization: Bearer $ROUTERPLEX_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "kimi-k3.1-pro", "messages": [{"role": "user", "content": "Hello!"}]}'

Questions

Before you pick one of the two.

Answers computed from the catalog
01Is DeepSeek cheaper than Kimi?+

At the flagship tier, DeepSeek V4 Pro costs $0.80 / $2.39 per 1M input / output tokens and Kimi K3.1 Pro costs $2.00 / $10.00. A coding-agent session comes to $0.90 on DeepSeek V4 Pro and $2.10 on Kimi K3.1 Pro. At the cheap tier, DeepSeek V4.1 Flash ($0.18 / $0.73) meets Kimi K2.6 ($0.95 / $4.00): $0.25 against $0.91 for the same session.

02Which DeepSeek models can I use on RouterPlex?+

DeepSeek V4.1 Flash (deepseek-v4.1-flash, $0.18 / $0.73), DeepSeek V4.1 Flash 1017 (deepseek-v4.1-flash-1017, $0.18 / $0.73), DeepSeek V4 Pro (deepseek-v4-pro, $0.80 / $2.39), DeepSeek V4 Pro 0813 (deepseek-v4-pro-0813, $0.80 / $2.39). This page treats DeepSeek V4 Pro as the DeepSeek flagship (the newest full-size version in the catalog) and DeepSeek V4.1 Flash as the cheap model (lowest cost on a 100K-input, 10K-output turn).

03Which Kimi models can I use on RouterPlex?+

Kimi K2.6 (kimi-k2.6, $0.95 / $4.00), Kimi K2.7 Code (kimi-k2.7-code, $0.95 / $4.00), Kimi K3.1 Pro (kimi-k3.1-pro, $2.00 / $10.00), Kimi K3 (kimi-k3, $3.00 / $15.00). This page treats Kimi K3.1 Pro as the Kimi flagship (the newest full-size version in the catalog) and Kimi K2.6 as the cheap model (lowest cost on a 100K-input, 10K-output turn).

04Does DeepSeek or Kimi have the larger context window?+

The flagships match at 1M: DeepSeek V4 Pro and Kimi K3.1 Pro both accept 1M tokens.

05Can I use DeepSeek and Kimi with the same API key?+

Yes. Every model on this page runs on one RouterPlex key through the OpenAI-compatible endpoint https://api.routerplex.com/v1 and the Anthropic-compatible /v1/messages. Switching between DeepSeek and Kimi is a one-line change of the model id, for example "deepseek-v4-pro" to "kimi-k3.1-pro".

Try both

Both models behind one key.

Top up from $5 with no subscription, then send the same prompt to each.