SPEED · COST · QUALITY
Measured matrix and verified gateway observations only.
- Overall (Matrix)
- 30.0
- Cadle Intelligence
- 7.2
Liquid AI · Open weightsLFM2.5-8B-A1B
Active
Model portfolio
LFM2.5-8B-A1B
Tap a star to rate
Liquid AI · Released 2026-05-28 · v2026-05-28 · Open
Language Cap
No data yet
language · unranked
Benchmark average
61.7
Not enough benchmark results yet
Pricing
Unlisted
no list price
CADLE CAP vs MODEL MATRIX
Cadle Cap ranks models on every published score we have for them — harder benchmarks weigh more. Missing tests are skipped, not zeroed. It answers who is stronger than whom in today's pool.
Model Matrix is the datasheet: Overall, Intelligence, Coding, Agentic, MMLU, GPQA, Math, LiveCodeBench, HumanEval, Reason, ARC-AGI. A dash means unmeasured. AVG is an importance-weighted mean of that model's filled cells only.
MODEL MATRIX · TRUTH CENTER CELLS
Overall
30.0
Intelligence
7.2
Coding
No data yet
Agentic
52.1
MMLU
46.9
GPQA
51.3
Math
88.8
LiveCodeBench
No data yet
HumanEval
No data yet
Reason
73.8
ARC-AGI
No data yet
Coverage AVG
61.7
Cadle Cap ranks models with within-benchmark percentiles so ARC-AGI ~40 and MMLU ~90 never compete raw. Coverage AVG is an importance-weighted mean of filled expert cells only — blanks are skipped, not zeroed.
ANALYSIS · TRUTH CENTER
Hard benches (GPQA · ARC-AGI · LCB · Agentic) sit left — a mid-40 on ARC-AGI can still be frontier. Broader benches on the right are easier and often higher.
NOT RANKED YET
We only rank a model once it has enough published benchmark results to compare fairly. Everything below is what we do know: pricing, news and any scores that have been published.
DESCRIPTION + DETAILS
PROVIDER
Liquid AI
RELEASE
2026-05-28
STATUS
ACTIVE
COMMUNITY REVIEWS
OPERATOR NOTES
Quality, latency, cost, and failure modes — verified production observations.
4.5
2 notes indexed
SUBMIT OPERATOR NOTE
Tap a star to rate
Topic:
5 - Frontier benchmark
Stored locally · synced to your profile
I
Inference Operator
VerifiedOct 23, 2026
Tested LFM2.5-8B-A1B on high-concurrency tool loops. Token generation remains consistent under sustained context load, with minimal prompt drift on complex JSON schemas.
A
Autonomous Agent Lead
VerifiedOct 19, 2026
Evaluated LFM2.5-8B-A1B on multi-turn code synthesis. Code reasoning and diff application succeed on ~88% of test suites. Best paired with strict system prompt guardrails.
MODEL COMPARISON
| Metric | |
|---|---|
| Provider | Liquid AI |
| Release | 2026-05-28 |
| Input / 1M | No data yet |
| Output / 1M | No data yet |
| Context | No data yet |
| Cadle Cap | 70 |
| Cap board | Unranked |
DATA PROVENANCE
Scores may use different scales — check the Benchmarks tab for source links.
- Supabase models
- Cadle public observations
MEASURED CELLS
Same Truth Center numbers as the datasheet above. Accuracy, reliability, safety, and cost scores are not invented.
Overall30
AA IQ7.2
Agentic52.1
MMLU46.9
GPQA51.3
Math88.8