Armand McQueen

basic-generation

ok

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

Produced 97 words (128 output tokens).

Result

{
  "wordCount": 97,
  "outputTokens": 128
}

1e129bba-72cd-434e-af4c-155daab64544

Sep 5, 2026, 02:35:23

google:gemini-3.8-flash

Input

14

Output

128

Reasoning

0

Answer

128

Cost

$0.000981

Mode

streaming

Cost breakdown

Input $0.000021Output $0.000960

Latency

1.9s

TTFT

1.3s

Tok/s

68.0

Tok/s (adj)

210.5

usage-cost

ok

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

in=10 out=152 rt=147 cost=$0.001155 (paid).

Result

{
  "inputTokens": 10,
  "outputTokens": 152,
  "reasoningTokens": 147,
  "costUsd": 0.001155,
  "paid": true
}

f9a38274-d63e-416a-b509-0e333a822610

Sep 5, 2026, 02:35:24

google:gemini-3.8-flash

Input

10

Output

152

Reasoning

147

Answer

5

Cost

$0.0012

Mode

streaming

Cost breakdown

Input $0.000015Output $0.0011

Latency

1.2s

TTFT

1.2s

Tok/s

122.3

Tok/s (adj)

4201.7

streaming-structure

ok

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

mode=streaming, 8 stream events, 1 text block(s), 35 chars.

Result

{
  "mode": "streaming",
  "streamEventCount": 8,
  "textBlocks": 1,
  "textLen": 35
}

eeebd689-059f-4b13-a960-57ce20c312cc

Sep 5, 2026, 02:35:25

google:gemini-3.8-flash

Input

15

Output

195

Reasoning

160

Answer

35

Cost

$0.0015

Mode

streaming

Cost breakdown

Input $0.000023Output $0.0015

Latency

1.0s

TTFT

996ms

Tok/s

190.6

Tok/s (adj)

1313.8

complete-stream-parity

ok

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

stream="Paris" · complete="Paris"

Result

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

084e9ea3-c274-40fb-a462-92bac9a1c780

Sep 5, 2026, 02:35:25

google:gemini-3.8-flash

Input

15

Output

1

Reasoning

0

Answer

1

Cost

$0.000030

Mode

streaming

Cost breakdown

Input $0.000023Output $0.000007

Latency

494ms

TTFT

493ms

Tok/s

2.0

Tok/s (adj)

729.9

1f2ce207-8e08-4b03-b6d5-56f2fd5fb8cb

Sep 5, 2026, 02:35:26

google:gemini-3.8-flash

Input

15

Output

1

Reasoning

0

Answer

1

Cost

$0.000030

Mode

non-streaming

Cost breakdown

Input $0.000023Output $0.000007

Latency

525ms

Tok/s

1.9

reasoning-conformance

ok

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

Class (from reasoning-config): always-on.
reasoningTokens=794, outputTokens=1197.

Result

{
  "class": "always-on",
  "reasoningTokens": 794,
  "outputTokens": 1197
}

3fc296bf-2812-4bbf-bcd8-c29326307201

Sep 5, 2026, 02:35:29

google:gemini-3.8-flash

Input

92

Output

1,197

Reasoning

794

Answer

403

Cost

$0.0091

Mode

streaming

Cost breakdown

Input $0.000138Output $0.0090

Latency

3.5s

TTFT

2.6s

Tok/s

341.2

Tok/s (adj)

423.6

single-tool-call

ok

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
}

a18381d7-d618-45ff-8fe9-285a40d9c66e

Sep 5, 2026, 02:35:30

google:gemini-3.8-flash

Input

99

Output

18

Reasoning

0

Answer

18

Cost

$0.000284

Mode

streaming

Cost breakdown

Input $0.000148Output $0.000135

Latency

709ms

Tok/s

25.4

tool-loop

ok

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

step1 (required): multiply({"a":23,"b":17}) → step2: "The product of 23 and 17 is 391."

Result

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

6259a4df-55ff-4de3-9772-b6c4089ae7f4

Sep 5, 2026, 02:35:31

google:gemini-3.8-flash

Input

95

Output

67

Reasoning

49

Answer

18

Cost

$0.000645

Mode

streaming

Cost breakdown

Input $0.000142Output $0.000503

Latency

901ms

Tok/s

74.4

00e7ca57-1deb-4a16-809f-0e7f54ea78b1

Sep 5, 2026, 02:35:32

google:gemini-3.8-flash

Input

182

Output

32

Reasoning

16

Answer

16

Cost

$0.000513

Mode

streaming

Cost breakdown

Input $0.000273Output $0.000240

Latency

606ms

TTFT

590ms

Tok/s

52.8

Tok/s (adj)

985.8

web-search

ok

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

ok

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

reasoned=true, google token captured; follow-up → "### Account Balance Lookup * **Account ID:** `acme-42` * **Current Bal"

Result

{
  "signatureType": "google",
  "tokenCaptured": true,
  "reasoned": true,
  "answered": true,
  "continuation": "### Account Balance Lookup\n* **Account ID:** `acme-42`\n* **Current Balance ($P$):** **$1,234.00**\n\n---\n\n### Growth Calculation\n* **Annual in"
}

a35bab9f-a75e-438c-8b47-75acbb584eac

Sep 5, 2026, 02:35:33

google:gemini-3.8-flash

Input

123

Output

111

Reasoning

91

Answer

20

Cost

$0.0010

Mode

streaming

Cost breakdown

Input $0.000185Output $0.000833

Latency

839ms

Tok/s

132.3

3927f2de-f2aa-4cba-b5df-129c58e124c4

Sep 5, 2026, 02:35:36

google:gemini-3.8-flash

Input

261

Output

999

Reasoning

657

Answer

342

Cost

$0.0079

Mode

streaming

Cost breakdown

Input $0.000392Output $0.0075

Latency

3.0s

TTFT

2.2s

Tok/s

327.9

Tok/s (adj)

415.8

prompt-cache

ok

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

implicit cache; cacheRead per attempt [4081]; hit=yes.

Result

{
  "cacheStyle": "implicit",
  "cacheReadPerAttempt": [
    4081
  ],
  "cacheReadTokens": 4081,
  "cacheHitCost": 0.01024215
}

228c49a7-67ed-40f2-8bc4-9dfcf3c893e7

Sep 5, 2026, 02:35:41

google:gemini-3.8-flash

Input

10,121

Output

77

Reasoning

77

Answer

0

Cost

$0.0158

Mode

non-streaming

Cost breakdown

Input $0.0152Output $0.000577

Latency

5.7s

Tok/s

13.5

baa0dd72-5c73-446b-a96f-ad5b0cc67f67

Sep 5, 2026, 02:35:48

google:gemini-3.8-flash

Input

10,121

Output

76

Reasoning

72

Answer

4

Cache Read

4,081

Cost

$0.0102

Mode

non-streaming

Cost breakdown

Input $0.0091Output $0.000570Cache read $0.000612

Latency

6.7s

Tok/s

11.3

json-output

ok

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

Output parsed as JSON.
All required keys present.
Output is bare JSON (no leading prose/fence).

Result

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

35ca2611-7498-4bc4-a623-02a161fc81d3

Sep 5, 2026, 02:35:49

google:gemini-3.8-flash

Input

57

Output

21

Reasoning

0

Answer

21

Cost

$0.000243

Mode

streaming

Cost breakdown

Input $0.000086Output $0.000157

Latency

924ms

TTFT

892ms

Tok/s

22.7

Tok/s (adj)

652.2

summarize-memex

ok

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

Read ~10155 input tokens, wrote 427 words (550 output tokens) in 3413ms.

Result

{
  "wordCount": 427,
  "inputTokens": 10155,
  "outputTokens": 550,
  "latencyMs": 3413
}

7dda9f10-8009-44b8-be34-32c066aad6f0

Sep 5, 2026, 02:35:52

google:gemini-3.8-flash

Input

10,155

Output

550

Reasoning

0

Answer

550

Cost

$0.0194

Mode

streaming

Cost breakdown

Input $0.0152Output $0.0041

Latency

3.4s

TTFT

869ms

Tok/s

161.1

Tok/s (adj)

216.2

truncation

ok

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

finishReason=length, out=96 tokens.

Result

{
  "finishReason": "length",
  "outputTokens": 96
}

219f976f-19cc-4771-a2f5-a20cbc087318

Sep 5, 2026, 02:35:53

google:gemini-3.8-flash

Input

21

Output

96

Reasoning

0

Answer

96

Cost

$0.000751

Mode

streaming

Cost breakdown

Input $0.000031Output $0.000720

Latency

930ms

TTFT

591ms

Tok/s

103.2

Tok/s (adj)

282.8

streaming-only-rejection

ok

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

streamingOnly=false; complete() succeeded.

Result

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

abort

ok

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

pre-aborted signal → rejected (AbortError).

Result

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