deepinfra:kimi-k3

8/17/2026, 7:43:08 AM · 15 tests · run e3961772-129a-40b9-83c9-f0c14bcf1b8b

basic-generation

Plain streaming generation with a soft length constraint (~100 words).

Produced 97 words (127 output tokens).

Result

{
  "wordCount": 97,
  "outputTokens": 127
}

da76a1e6-5fc7-4b84-b4c0-16213b3445ce

Aug 17, 2026, 07:41:59

deepinfra:moonshotai/Kimi-K3

Input

Output

127

Reasoning

Answer

127

Cost

$0.0019

Mode

streaming

Cost breakdown

Input $0.000091Output $0.0018

Latency

3.9s

TTFT

618ms

Tok/s

32.5

Tok/s (adj)

38.6

usage-cost

Asserts usage tokens populate and cost computes from the ChatReq contract.

in=96 out=203 rt=0 cost=$0.003166 (paid).

Result

{
  "inputTokens": 96,
  "outputTokens": 203,
  "reasoningTokens": 0,
  "costUsd": 0.00316635,
  "paid": true
}

d72d47dc-c066-4f2f-aff4-24e7f85a86e6

Aug 17, 2026, 07:42:04

deepinfra:moonshotai/Kimi-K3

Input

Output

203

Reasoning

Answer

203

Cost

$0.0032

Mode

streaming

Cost breakdown

Input $0.000274Output $0.0029

Latency

5.1s

TTFT

4.8s

Tok/s

40.1

Tok/s (adj)

833.9

streaming-structure

Asserts a streaming call captures events that reconstruct into the final text.

mode=streaming, 57 stream events, 1 text block(s), 35 chars.

Result

{
  "mode": "streaming",
  "streamEventCount": 57,
  "textBlocks": 1,
  "textLen": 35
}

592000da-451a-4900-8d83-69bbfe4d0838

Aug 17, 2026, 07:42:07

deepinfra:moonshotai/Kimi-K3

Input

101

Output

Reasoning

Answer

Cost

$0.0012

Mode

streaming

Cost breakdown

Input $0.000288Output $0.000884

Latency

2.6s

TTFT

1.1s

Tok/s

24.3

Tok/s (adj)

43.9

complete-stream-parity

Same prompt via stream() and complete() both return valid, matching results.

stream="Paris" · complete="Paris"

Result

{
  "streamText": "Paris",
  "completeText": "Paris"
}

0c47f446-926b-4157-a682-7eb298ddb4b3

Aug 17, 2026, 07:42:08

deepinfra:moonshotai/Kimi-K3

Input

Output

Reasoning

Answer

Cost

$0.000214

Mode

streaming

Cost breakdown

Input $0.000100Output $0.000114

Latency

592ms

TTFT

412ms

Tok/s

13.5

Tok/s (adj)

44.5

411f3cda-42e7-45cb-8baf-12566fdd5a33

Aug 17, 2026, 07:42:08

deepinfra:moonshotai/Kimi-K3

Input

Output

Reasoning

Answer

Cost

$0.000214

Mode

non-streaming

Cost breakdown

Input $0.000100Output $0.000114

Latency

526ms

Tok/s

15.2

reasoning-conformance

Asserts the model's reasoning knob behaves as its reasoning-config declares.

Class (from reasoning-config): toggleable.
off: out=506 · mid: out=356 (rt=0, rchars=311).

Result

{
  "class": "toggleable",
  "mode": "deterministic",
  "offOutputTokens": 506,
  "midOutputTokens": 356,
  "midReasoningTokens": 0
}

df95ffe6-07f5-4eac-bfa4-c00f45acb9c2

Aug 17, 2026, 07:42:20

deepinfra:moonshotai/Kimi-K3

Input

106

Output

506

Reasoning

Answer

506

Cost

$0.0075

Mode

streaming

Cost breakdown

Input $0.000302Output $0.0072

Latency

12.2s

TTFT

215ms

Tok/s

41.4

Tok/s (adj)

42.2

9d5aca5f-6508-4ae2-85bd-8cb34cd3ce6a

Aug 17, 2026, 07:42:31

deepinfra:moonshotai/Kimi-K3

Input

174

Output

356

Reasoning

Answer

356

Cost

$0.0056

Mode

streaming

Cost breakdown

Input $0.000496Output $0.0051

Latency

10.3s

TTFT

3.9s

Tok/s

34.5

Tok/s (adj)

55.6

single-tool-call

Offers a calculator tool and checks the model calls it for an arithmetic question.

Called "multiply" with args {"a":23,"b":17}.

Result

{
  "toolCalled": true,
  "toolName": "multiply",
  "args": {
    "a": 23,
    "b": 17
  },
  "correct": true
}

06538f5c-7879-4cf6-bd3d-b389d346d17f

Aug 17, 2026, 07:42:32

deepinfra:moonshotai/Kimi-K3

Input

175

Output

Reasoning

Answer

Cost

$0.0013

Mode

streaming

Cost breakdown

Input $0.000499Output $0.000812

Latency

1.7s

Tok/s

34.4

tool-loop

Forces a tool call, feeds the result back, and checks the continuation uses it.

step1 (required): multiply({"a":23,"b":17}) → step2: "23 multiplied by 17 is 391."

Result

{
  "toolName": "multiply",
  "args": {
    "a": 23,
    "b": 17
  },
  "product": 391,
  "continuation": "23 multiplied by 17 is 391.",
  "usesResult": true,
  "forced": true
}

b9fd4818-dfa0-4b61-8c3c-4ebb9251d3c5

Aug 17, 2026, 07:42:35

deepinfra:moonshotai/Kimi-K3

Input

278

Output

Reasoning

Answer

Cost

$0.0020

Mode

streaming

Cost breakdown

Input $0.000792Output $0.0012

Latency

2.7s

Tok/s

31.6

efc19eb9-1a19-404d-8c9a-fa2581e0a193

Aug 17, 2026, 07:42:36

deepinfra:moonshotai/Kimi-K3

Input

363

Output

Reasoning

Answer

Cost

$0.0014

Mode

streaming

Cost breakdown

Input $0.0010Output $0.000370

Latency

1.0s

TTFT

675ms

Tok/s

25.8

Tok/s (adj)

78.4

web-search

Enables server-side web search and checks invocation counts, the per-request cost surcharge, and provider-executed output blocks.

Skipped: model does not support the catalog's web_search tool version.

Result

{
  "skipped": "no server-side web search support"
}

reasoning-tool-continuity

Reasoning model: continuity token captured on a tool call and replayed so the follow-up succeeds.

Step 1 didn't engage reasoning (adaptive discretion) — signature not exercised this run.
reasoned=false, none token —; follow-up → "The balance for account `acme-42` is $1,234. After 7 years at 5% comp"

Result

{
  "signatureType": "none",
  "tokenCaptured": false,
  "reasoned": false,
  "answered": true,
  "continuation": "The balance for account `acme-42` is $1,234.\n\nAfter 7 years at 5% compounded annually:\n$1,234 × (1.05)^7 ≈ $1,736.36, rounded to **$1,736**."
}

a39d6174-390f-4a88-a804-98883bdea725

Aug 17, 2026, 07:42:45

deepinfra:moonshotai/Kimi-K3

Input

276

Output

Reasoning

Answer

Cost

$0.0020

Mode

streaming

Cost breakdown

Input $0.000787Output $0.0012

Latency

2.9s

Tok/s

28.5

15e0ebbe-8680-4bed-b541-94bf5a059ea7

Aug 17, 2026, 07:42:48

deepinfra:moonshotai/Kimi-K3

Input

292

Output

Reasoning

Answer

Cost

$0.0019

Mode

streaming

Cost breakdown

Input $0.000832Output $0.0011

Latency

2.9s

TTFT

1.1s

Tok/s

27.1

Tok/s (adj)

43.4

prompt-cache

Repeat a large prefix; assert cache-read tokens are reported and cost drops.

Skipped: provider has no (known) prompt-cache reporting.

Result

{
  "skipped": "cacheStyle=none"
}

json-output

Requests a strict JSON object and checks it parses with the required keys.

Output parsed as JSON.
All required keys present.
Output had leading prose or a code fence before the JSON.

Result

{
  "parsed": {
    "name": "Anthropic",
    "founded": 2021,
    "headquarters": "San Francisco, California, United States"
  },
  "validJson": true,
  "missingKeys": [],
  "bareJson": false
}

fdb532db-33e1-43bf-8e46-a97867e0d2e3

Aug 17, 2026, 07:42:49

deepinfra:moonshotai/Kimi-K3

Input

Output

Reasoning

Answer

Cost

$0.000898

Mode

streaming

Cost breakdown

Input $0.000257Output $0.000641

Latency

1.2s

TTFT

255ms

Tok/s

37.0

Tok/s (adj)

46.8

summarize-memex

Summarizes an ~11k-token essay to ~500 words (long input, reasoning minimized).

Read ~10277 input tokens, wrote 474 words (622 output tokens) in 15213ms.

Result

{
  "wordCount": 474,
  "inputTokens": 10277,
  "outputTokens": 622,
  "latencyMs": 15213
}

242451dd-8551-4dcf-a719-42cc58d82669

Aug 17, 2026, 07:43:04

deepinfra:moonshotai/Kimi-K3

Input

10,277

Output

622

Reasoning

Answer

622

Cache Read

9,216

Cost

$0.0145

Mode

streaming

Cost breakdown

Input $0.0030Output $0.0089Cache read $0.0026

Latency

15.2s

TTFT

851ms

Tok/s

40.9

Tok/s (adj)

43.3

truncation

Tiny maxTokens on a long prompt yields finishReason:"length".

finishReason=length, out=100 tokens.

Result

{
  "finishReason": "length",
  "outputTokens": 100
}

407bb2c6-f2e3-4bfc-990d-7f177c41773e

Aug 17, 2026, 07:43:07

deepinfra:moonshotai/Kimi-K3

Input

Output

100

Reasoning

Answer

100

Cost

$0.0015

Mode

streaming

Cost breakdown

Input $0.000108Output $0.0014

Latency

2.8s

TTFT

326ms

Tok/s

35.3

Tok/s (adj)

39.9

streaming-only-rejection

Streaming-only models reject complete(); all others accept it.

streamingOnly=false; complete() succeeded.

Result

{
  "streamingOnly": false,
  "completeThrew": false
}

abort

A pre-aborted signal rejects the call (abort is threaded + honored).

pre-aborted signal → rejected (AbortError).

Result

{
  "rejected": true,
  "errorName": "AbortError"
}