deepinfra:nemotron-3-super

8/17/2026, 7:44:55 AM · 15 tests · run 615689ef-0103-4182-85c9-774b9d080dae

basic-generation

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

Produced 19 words (26 output tokens).

Result

{
  "wordCount": 19,
  "outputTokens": 26
}

8cc97f42-2b3a-4aaa-8fbf-db3505637a93

Aug 17, 2026, 07:43:19

deepinfra:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B

Input

Output

Reasoning

Answer

Cost

$0.000013

Mode

streaming

Cost breakdown

Input $0.000002Output $0.000010

Latency

1.2s

TTFT

843ms

Tok/s

22.1

Tok/s (adj)

78.0

usage-cost

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

in=25 out=74 rt=0 cost=$0.000032 (paid).

Result

{
  "inputTokens": 25,
  "outputTokens": 74,
  "reasoningTokens": 0,
  "costUsd": 0.000031725,
  "paid": true
}

07144f36-b326-4973-a8f2-f21411844f76

Aug 17, 2026, 07:43:20

deepinfra:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B

Input

Output

Reasoning

Answer

Cost

$0.000032

Mode

streaming

Cost breakdown

Input $0.000002Output $0.000030

Latency

1.6s

TTFT

1.5s

Tok/s

44.8

Tok/s (adj)

505.3

streaming-structure

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

mode=streaming, 62 stream events, 1 text block(s), 63 chars.

Result

{
  "mode": "streaming",
  "streamEventCount": 62,
  "textBlocks": 1,
  "textLen": 63
}

457cb8f7-313e-400f-a32d-0866c06bf115

Aug 17, 2026, 07:43:27

deepinfra:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B

Input

Output

460

Reasoning

Answer

460

Cost

$0.000187

Mode

streaming

Cost breakdown

Input $0.000003Output $0.000184

Latency

6.7s

TTFT

6.0s

Tok/s

69.0

Tok/s (adj)

705.1

complete-stream-parity

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

stream="Paris" · complete="Paris"

Result

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

64268f3d-32ec-4cb9-b8e8-2ab49dafe3f8

Aug 17, 2026, 07:43:27

deepinfra:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B

Input

Output

Reasoning

Answer

Cost

$0.000003

Mode

streaming

Cost breakdown

Input $0.000003Output $0.000001

Latency

422ms

TTFT

422ms

Tok/s

4.7

Tok/s (adj)

5128.2

67557117-5e50-4606-b811-35534bd6bcba

Aug 17, 2026, 07:43:28

deepinfra:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B

Input

Output

Reasoning

Answer

Cost

$0.000003

Mode

non-streaming

Cost breakdown

Input $0.000003Output $0.000001

Latency

317ms

Tok/s

6.3

reasoning-conformance

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

Class (from reasoning-config): toggleable.
off: out=814 · mid: out=920 (rt=0, rchars=788).

Result

{
  "class": "toggleable",
  "mode": "deterministic",
  "offOutputTokens": 814,
  "midOutputTokens": 920,
  "midReasoningTokens": 0
}

b6618e6e-2c40-4c8c-9f65-0e6f8212905a

Aug 17, 2026, 07:43:48

deepinfra:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B

Input

108

Output

814

Reasoning

Answer

814

Cost

$0.000335

Mode

streaming

Cost breakdown

Input $0.000009Output $0.000326

Latency

20.0s

TTFT

451ms

Tok/s

40.8

Tok/s (adj)

41.7

e40b5aed-d125-465c-863e-5d4ddedd2c34

Aug 17, 2026, 07:44:02

deepinfra:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B

Input

108

Output

920

Reasoning

Answer

920

Cost

$0.000377

Mode

streaming

Cost breakdown

Input $0.000009Output $0.000368

Latency

14.2s

TTFT

5.1s

Tok/s

64.7

Tok/s (adj)

100.8

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
}

585f8c2a-c450-479d-848c-822118785b0f

Aug 17, 2026, 07:44:03

deepinfra:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B

Input

366

Output

Reasoning

Answer

Cost

$0.000045

Mode

streaming

Cost breakdown

Input $0.000031Output $0.000014

Latency

765ms

TTFT

423ms

Tok/s

45.8

Tok/s (adj)

102.4

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 equals 391."

Result

{
  "toolName": "multiply",
  "args": {
    "a": 23,
    "b": 17
  },
  "product": 391,
  "continuation": "\n\n23 multiplied by 17 equals 391.",
  "usesResult": true,
  "forced": true
}

83b5ce1e-883f-42c5-a4ca-dc391fd274d2

Aug 17, 2026, 07:44:06

deepinfra:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B

Input

360

Output

Reasoning

Answer

Cost

$0.000061

Mode

streaming

Cost breakdown

Input $0.000031Output $0.000030

Latency

3.7s

TTFT

1.1s

Tok/s

20.3

Tok/s (adj)

28.6

c9a91f52-c886-4665-94e0-fe1f40540125

Aug 17, 2026, 07:44:07

deepinfra:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B

Input

458

Output

Reasoning

Answer

Cost

$0.000057

Mode

streaming

Cost breakdown

Input $0.000039Output $0.000018

Latency

901ms

TTFT

746ms

Tok/s

48.8

Tok/s (adj)

283.8

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 → "Current balance: **$1,234 USD** (from `acme-42`). **Growth calculatio"

Result

{
  "signatureType": "none",
  "tokenCaptured": false,
  "reasoned": false,
  "answered": true,
  "continuation": "Current balance: **$1,234 USD** (from `acme-42`).\n\n**Growth calculation (5 % per year, compounded annually for 7 years):**  \n\n\\[\n\\text{Futur"
}

4938f980-5012-4419-a46f-18ca58c11dfe

Aug 17, 2026, 07:44:15

deepinfra:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B

Input

376

Output

124

Reasoning

Answer

124

Cost

$0.000082

Mode

streaming

Cost breakdown

Input $0.000032Output $0.000050

Latency

4.2s

TTFT

3.7s

Tok/s

29.6

Tok/s (adj)

270.9

86921bea-02e9-4aac-b72a-a0385073ca58

Aug 17, 2026, 07:44:27

deepinfra:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B

Input

527

Output

303

Reasoning

Answer

303

Cost

$0.000166

Mode

streaming

Cost breakdown

Input $0.000045Output $0.000121

Latency

11.4s

TTFT

613ms

Tok/s

26.5

Tok/s (adj)

28.0

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 is bare JSON (no leading prose/fence).

Result

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

6ef1228e-9cf7-40d7-8c3b-51b844b2ff5a

Aug 17, 2026, 07:44:28

deepinfra:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B

Input

Output

Reasoning

Answer

Cost

$0.000021

Mode

streaming

Cost breakdown

Input $0.000006Output $0.000014

Latency

1.4s

TTFT

437ms

Tok/s

26.1

Tok/s (adj)

38.3

summarize-memex

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

Read ~10385 input tokens, wrote 467 words (627 output tokens) in 21473ms.

Result

{
  "wordCount": 467,
  "inputTokens": 10385,
  "outputTokens": 627,
  "latencyMs": 21473
}

b3b54d94-4736-48e4-add1-f44dab1cea0d

Aug 17, 2026, 07:44:50

deepinfra:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B

Input

10,385

Output

627

Reasoning

Answer

627

Cost

$0.0011

Mode

streaming

Cost breakdown

Input $0.000883Output $0.000251

Latency

21.5s

TTFT

632ms

Tok/s

29.2

Tok/s (adj)

30.1

truncation

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

finishReason=length, out=100 tokens.

Result

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

26969c80-a45c-4bbd-893e-7ab5afea9717

Aug 17, 2026, 07:44:53

deepinfra:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B

Input

Output

100

Reasoning

Answer

100

Cost

$0.000043

Mode

streaming

Cost breakdown

Input $0.000003Output $0.000040

Latency

3.1s

TTFT

734ms

Tok/s

32.7

Tok/s (adj)

43.1

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"
}