⌘K
—

AI-Generated Recommendations

Findings EnsembleAI surfaced across evaluation, optimization, and reliability — triage them here.

New

4

awaiting triage

In review

5

Actioned

4

all time
In ReviewLatencySolis Pro 4.1

Regional latency gap widening between EU and APAC

Solis Pro 4.1 p95 latency in APAC is now 2.1x the EU figure, up from 1.4x a month ago.

EnsembleAI97% confidence16h ago
NewLatencySolis Flash 4.1

Latency regression detected for Solis Flash 4.1 in ap-southeast-1

p99 latency is up 34% over the last 6 hours in this region alone. Investigate the caching layer for that cluster.

EnsembleAI97% confidence2d ago
DismissedQualitySolis Motion

Self-evaluation scores for Solis Motion trending upward

The last 5 self-eval runs show a consistent upward trend across Reasoning-500 and Multimodal-QA.

EnsembleAI68% confidence3d ago
In ReviewQualitySolis Code 2.0

Solis Code 2.0 completions show declining acceptance in TypeScript

Acceptance rate for TypeScript completions dropped 9 points this week versus other languages.

EnsembleAI79% confidence4d ago
In ReviewCapabilitySolis Ultra 4.1

API partners repeatedly requesting higher output token limits

Six enterprise partners have filed capability requests for output above the current maximum this month.

EnsembleAI71% confidence5d ago
ActionedReliabilityMumbai

Mumbai data center trending toward thermal threshold

Rack temperatures have climbed steadily for 48 hours. Recommend rebalancing load to nearby regions before it forces a throttle.

EnsembleAI84% confidence5d ago
NewCostToronto

Underutilized reserved capacity in Toronto

Average load has held at 38% for three weeks — well below the reservation. Consider resizing.

EnsembleAI84% confidence7d ago
ActionedCostSolis Ultra 3.5

Solis Ultra 3.5 traffic has fallen below the dedicated-capacity threshold

Daily requests are down to 210M, low enough to fold serving into the shared 4.1 pool.

EnsembleAI96% confidence8d ago
ActionedLatencySolis Motion 1.2

Video generation queue wait time exceeds target during peak hours

Three regions saw queue wait times above the internal SLA during yesterday's evening peak.

EnsembleAI78% confidence9d ago
In ReviewReliabilityHuman Feedback Queue

Human feedback queue backlog growing

Submissions are up 40% week over week while grading throughput has stayed flat.

EnsembleAI75% confidence9d ago

Showing 1–10 of 14 recommendations

10
per page1of 2