tinyanalysis

Leaderboard / ServiceNow (Apriel)

Apriel 1.6 15B Thinker

ServiceNow (Apriel) · Apriel · released 2025-12-04 · data checked 2026-11-01

Apriel 1.6 15B Thinker is a 14.9B-parameter dense model from ServiceNow (Apriel) (United States), released 2025-12-04 under the MIT license. It scores 60.9 on the Small LM Index, #3 of 26 small models. It needs about 11.4 GB of memory at 4-bit with an 8K-token context.

Small LM Index
60.9
#3 of 26 scored small models
Likely 59.1–62.6 · from 4 of 6 core benchmarks
Parameters
14.9B
Dense
Price per 1M tokens
—
No hosted API on OpenRouter; self-host
Context window
262K
tokens

Benchmarks

Each bar shows this model's score. The tick marks the best small model on that benchmark. Sources: 9 reported by ServiceNow (Apriel). How we label sources.

MMLU-Pro Lab-reported Knowledge79.0
Class best: Qwen3.5 9B 82.5
GPQA Diamond Lab-reported Knowledge73.0
Class best: Qwen3.5 9B 81.7
Humanity's Last Exam Lab-reported Reasoning10.0
Class best: Falcon H1R 7B 11.1
AIME Lab-reported Math · AIME 202588.0
Terminal-Bench Lab-reported Agentic · Terminal-Bench (version not stated)14.0
Class best: Granite 4.2 8B 20.6
τ²-Bench Lab-reported Agentic · Average of Telecom, Retail and Airline64.6
Class best: Qwen3.5 9B 79.1
MMMU-Pro Lab-reported Vision60.3
Class best: Qwen3.5 9B 69.2
BFCL Lab-reported Agentic · BFCL v363.5
Class best: Hunyuan 7B Instruct 70.8
AA-LCR Lab-reported Long context50.0

Not reported: estimated

What the index model expects from this model's other scores. Not a measurement.

LiveCodeBench≈ 70
SWE-bench Verified≈ 47

Specifications

Lab
ServiceNow (Apriel) (US · United States)
Architecture
Dense transformer
Total parameters
14.9B
Active parameters
All (dense)
Size class
Small
Context window
262K tokens
Input modalities
text, image
Reasoning
Reasoning (thinking)
License
MIT
Openness
Open weights
Memory to run
~11.4 GB at 4-bit, 8K context (more sizes)
Hugging Face
1.2K downloads (30 days) · 308 likes
Weights on Hugging Face ↗

Compare with

Ranked in

Sources

Something wrong here? Report it.

Run it locally

Uses ServiceNow-AI/Apriel-1.6-15b-Thinker-GGUF (Q4_K_M). Swap the tag for another size: Q4_K_M.

Ollama
ollama run hf.co/ServiceNow-AI/Apriel-1.6-15b-Thinker-GGUF:Q4_K_M
llama.cpp
llama-server -hf ServiceNow-AI/Apriel-1.6-15b-Thinker-GGUF:Q4_K_M
Then open http://localhost:8080 for a chat window.
vLLM (GPU server)
vllm serve ServiceNow-AI/Apriel-1.6-15b-Thinker
Full-precision weights, for a server GPU.

Open in LM Studio

Memory needed: weights, context cache and runtime buffers.

Quantization4K context32K context256K context
16-bit F1631 GB37 GB82 GB
8-bit Q8_017 GB23 GB68 GB
4-bit Q4_K_M11 GB16 GB61 GB

Check it against your GPU, Mac or phone →