🧪 CLAUDE LAB · ANTHROPIC · OPUS 4.6 · SONNET 4.6 · HAIKU 4.5 · SWE-BENCH · SAFE TOOLING · CODE · VISION · COST 🧪 CLAUDE LAB · ANTHROPIC · OPUS 4.6 · SONNET 4.6 · HAIKU 4.5 · SWE-BENCH · SAFE TOOLING · CODE · VISION · COST

Anthropic runtime

Claude Grid

Model telemetry, benchmark interpretation, and operational health in one futuristic control surface.

LIVE
Claude in AI Race All Labs

Top model

Claude Opus 4.6

Flagship

MMLU 90.5 · HE 80.8 · Math 95

Latency 140ms · Context 1M

Live RPS

8.4k

tokens today 1.1B

Queue depth 9 · Avg latency 84ms

Active users

2.6M

Uptime 99.62%

Quality 88 · Value 234.0

Best value model

Claude Mini

Compact

Value 290.0 · io 3.2x

Speed lead: Claude Mini

Cheapest

$0.25 / $0.8

per 1M tokens in / out

ratio 3.2x · from Claude Mini

Context champion

Claude Opus 4.6

1M

HE Claude Sonnet 4.6 · Math Claude Opus 4.6

Model board

Current ranking: Quality score.

Claude Sonnet 4.6

Balanced

sonnet

Strong daily-driver model with resilient tool-calling behavior at scale.

MMLU

90.2

Latency 95ms

Context 1M

HumanEval 85.4

Math 92.6

Cost $3/$15

IO ratio 5x

Quality 89.0

Value 261.0

Use: Stable · Production · Coding · Vision

Claude Opus 4.6

Flagship

opus

Deep verification tasks, long-form drafting, and orchestration work.

MMLU

90.5

Latency 140ms

Context 1M

HumanEval 80.8

Math 95

Cost $5/$25

IO ratio 5x

Quality 88.0

Value 234.0

Use: Reasoning · Long Context · Code · Safety

Claude Haiku 4.5

Cost-Efficient

haiku

Great for high-volume loops where response speed and spend control matter.

MMLU

88.8

Latency 60ms

Context 200K

HumanEval 83.1

Math 89.1

Cost $0.8/$4

IO ratio 5x

Quality 86.9

Value 279.8

Use: Low Cost · Throughput · Multitask

Claude Mini

Compact

mini

Built for scale jobs where throughput and margin dominate the budget.

MMLU

86.7

Latency 42ms

Context 128K

HumanEval 77.8

Math 84

Cost $0.25/$0.8

IO ratio 3.2x

Quality 83.0

Value 290.0

Use: Batch · Automation · Summarization

Use-case routing map

Recommended model by product scenario.

SWE-bench / Tooling

Opus 4.6

Highest reasoning consistency for verification and refactor workflows.

Agent Workloads

Sonnet 4.6

Balanced quality and speed with stable tool orchestration behavior.

High Throughput

Mini

Lowest latency and best economics for background and batch runs.

Creative + Vision

Haiku 4.5

Fast multimodal responses with clean structure and readability.

⚡ Live pulse

sample stream
  1. Claude Opus 4.6 1m ago

    600k QPS · $5 in / $25 out · lat 140ms · ctx 1M · speed 50

  2. Claude Sonnet 4.6 2m ago

    950k QPS · $3 in / $15 out · lat 95ms · ctx 1M · speed 75

  3. Claude Haiku 4.5 3m ago

    1276k QPS · $0.8 in / $4 out · lat 60ms · ctx 200K · speed 98

  4. Claude Mini 4m ago

    1590k QPS · $0.25 in / $0.8 out · lat 42ms · ctx 128K · speed 120

🛰️ Anthropic ecosystem

—

Resolving GitHub repo health…

Claude model status

live

Claude Opus 4.6

STEADY

Uptime 99.14%

Latency 136ms

RPS 2.3k

Error % 0.02

Throughput 54k

Stream vision

Quality 88

Context 1M

Claude Sonnet 4.6

STEADY

Uptime 99.02%

Latency 99ms

RPS 2.8k

Error % 0.05

Throughput 79k

Stream vision

Quality 89

Context 1M

Claude Haiku 4.5

STEADY

Uptime 99.48%

Latency 58ms

RPS 3.2k

Error % 0.02

Throughput 97k

Stream vision

Quality 86.9

Context 200K

Claude Mini

STEADY

Uptime 99.59%

Latency 38ms

RPS 3.6k

Error % 0.03

Throughput 114k

Stream text

Quality 83

Context 128K

Throughput graph

Queue depth: 9