Skip to content
OSMEXO

Granite 4.0 Micro

ibm/granite-4.0-h-micro
  • Ibm
  • Family · Granite 4 Micro
  • Released 22 Sep 2025
  • Open weights
  • Text → Text
Not on OSMEXO yet
Spec & list price
Context128K
Max output118K
Input / 1M$0.02
Output / 1M$0.116.6× input
Scores · standing
Intelligence5.1#459/490
Usage
Requests / day42.1K
Per request · 30 days
Avg input1.5K0.1× median
Avg output1430.2× median

Consumption

133.3M266.7M400.0M27 Sep11 Oct26 Oct
Peak 380.3M 27 SepLow 66.0M 26 Oct

Benchmarks

bar = share of the top score · Benchmark data: Artificial Analysis
MMLU-Pro#238 of 273
44.7%
GPQA Diamond#387 of 445
33.6%
IFBench#300 of 328
24.8%
LiveCodeBench#208 of 269
18.0%
τ²-Bench#278 of 319
12.6%
Intelligence index#459 of 490
5.1
Humanity's Last Exam#328 of 465
5.0%
Long-context reasoning#327 of 393
7.0%
AIME 2025#183 of 206
6.0%
Math index#183 of 206
6.0
Terminal-Bench Hard#248 of 311
1.5%

Best value in its score band

Value
Intelligence Index
46810$0.02$0.05$0.1Granite 4.0 Micro
Blended price per 1M tokens (log)

Technical details

Tokenizer
Other
License
Apache 2.0
Parameters
3.2B
Moderation
none
Downloads · likes
8.5K · 148
Request parameters 14
  • response_format
  • temperature
  • top_p
  • top_k
  • min_p
  • max_tokens
  • stop
  • seed
  • frequency_penalty
  • logit_bias
+4 more
  • logprobs
  • presence_penalty
  • repetition_penalty
  • top_logprobs

When to use

Best for

  • Low-cost bulk text classification
  • Summarizing long documents cheaply
  • Self-hosted enterprise assistants (open weights)
  • Instruction-following in small pipelines
  • Retrieval-augmented question answering

Think twice for

  • Hardest reasoning and coding tasks
  • Tool-calling agents
  • Image understanding
  • Long autonomous agent runs

Alternatives

other vendors · similar capability

Want Granite 4.0 Micro on OSMEXO?

Sign in to request it. Requests are reviewed; a request does not guarantee availability.