CADLE CAP
53
Cadle #329/385 · 70% confidence
Google · Open weightsGemma 4 E4B
Active
Model portfolio
Gemma 4 E4B
Tap a star to rate
Google · Released 2026-04-03 · v2026-04-03 · Open
Gemma 4 E4B is an open-source model from Google (released 2026-04-03). Context window size is not listed with capabilities: Text, Code, Reasoning. List pricing ≈ $0.02 / $0.10 per 1M input/output tokens. Top tracked signals: MMLU · benchlm 69.4, MMLU 69.4, REASON · benchlm 60.0. Cadle merges provider listings, unified leaderboard medians, and public pricing where available.
Language Cap
53
Cadle Language #329/385 · 70%
Benchmark average
39.9
#394 on the benchmark matrix
Pricing
$0.020
$0.10 / 1M out · 220.0 pts/$
CROSS-BOARD RANKS
Cadle Language LLM #329/385Matrix #394/443BenchLM #181/224AA #449/698vs BenchLM: +132
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
30.9
AA IQ
8.9
Coding
12.5
Agentic
20.8
MMLU
69.4
GPQA
58.1
Math
—
LiveCodeBench
38.2
HumanEval
—
Reason
52.1
ARC-AGI
—
Coverage AVG
39.9
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.
CADLE INSIGHTS
Only measured signals
INTELLIGENCE
16.8
Media Cap quality axis
CODING
12.5
Media Cap quality axis
AGENTIC
20.8
Media Cap quality axis
KNOWLEDGE
62.0
Media Cap quality axis
MATRIX OVERALL
30.9
AVG 39.9 across matrix cells
BENCHMARK AVG
37.6
20 tracked scores
STRONGEST SIGNAL
MMLU · benchlm
Score 69.4
SPEED · COST · QUALITY
Measured matrix / AA / gateway rows only.
- Overall (Matrix)
- 30.9
- Cap Intelligence
- 26.4
- AA / Intel index
- 8.9
- Cap Coding
- 7.7
- Matrix Coding
- 12.5
- Speed tok/s
- 49
- Latency
- 375 ms
- Input $/1M
- $0.020
- Output $/1M
- $0.10
- Value pts/$
- 220.00
HUB · PARAMS
Hugging Face / parameter facts when mapped.
- Params~4B
Source: Hugging Face Hub
DESCRIPTION + DETAILS
Gemma 4 E4B is an open-source model from Google (released 2026-04-03). Context window size is not listed with capabilities: Text, Code, Reasoning. List pricing ≈ $0.02 / $0.10 per 1M input/output tokens. Top tracked signals: MMLU · benchlm 69.4, MMLU 69.4, REASON · benchlm 60.0. Cadle merges provider listings, unified leaderboard medians, and public pricing where available.
PROVIDER
RELEASE
2026-04-03
PARAMETERS
~4B
STATUS
ACTIVE
COMMUNITY REVIEWS
OPERATOR NOTES
Quality, latency, cost, and failure modes — not marketing copy.
No reviews yet
WRITE A REVIEW
Tap a star to rate
Stored on this device · synced when signed in
No operator notes yet
First review sets the tone for how teams evaluate Gemma 4 E4B.
MODEL COMPARISON
| Metric | |
|---|---|
| Provider | |
| Release | 2026-04-03 |
| Input / 1M | $0.02 |
| Output / 1M | $0.10 |
| Context | — |
| Cadle Cap | 53 |
| Cap board | Cadle Language #329/385 |
| MMLU · benchlm | 69.4 |
| MMLU | 69.4 |
| REASON · benchlm | 60.0 |
| GPQA · benchlm | 58.6 |
| GPQA | 58.1 |
| GPQA · artificial-analysis | 57.6 |
| REASON | 52.1 |
| LIVECODEBENCH · benchlm | 52.0 |
| REASON · artificial-analysis | 44.2 |
| LIVECODEBENCH | 38.2 |
| OVERALL · benchlm | 30.9 |
| OVERALL | 30.9 |
DATA PROVENANCE
Scores may use different scales — check the Benchmarks tab for source links.
- OpenRouter / provider APIs (pricing & context)
- Hugging Face Hub (downloads & cards)
- Papers With Code (task benchmarks)
- BenchLM.ai (category rollups + atomic benches)
- Epoch AI Benchmarking Hub (ECI + ZIP CSVs)
- LiveCodeBench (generation / execution / repair / test-output)
- Artificial Analysis (intelligence / indexes)
- LLM Stats (leaderboard + rankings)
- Cadle unified median (equal-weight multi-source)
- Artificial Analysis — throughput, latency, and $/M token snapshots
- Epoch AI Benchmarking Hub — capability CSV exports (when synced)
SOURCE TRACKING
Evidence linked to this model — listing, hub, benchmarks, or live gateway prices.
API
Google listing
API
artificial-analysis
API
Hugging Face Hub
API
OpenRouter
COMMUNITY
Papers with Code
COMMUNITY
BenchLM
COMMUNITY
unified-median
MEASURED CELLS
Same Truth Center numbers as the datasheet above. Accuracy, reliability, safety, and cost scores are not invented.
Overall30.9
AA IQ8.9
Coding12.5
Agentic20.8
MMLU69.4
GPQA58.1
LiveCodeBench38.2
AA Index8.9