Updated minutes ago
ReleasedDecember 15, 2025Verified 3d ago · docs.anthropic.com
Anthropic

Claude Opus 4.6

Anthropic · moe · 450B parameters · 1,000,000 context

Quality
90.0
Deploy this model →Use via API →Calculate ROI →Compare with othersFine-tune this model

Parameters

450B

Context Window

977K tokens

Architecture

MoE

Best GPU

B200 NVL (pair)

Cheapest API

$25.00/M

Quality Score

90/100

Intelligence Brief

Claude Opus 4.6 is a 450B parameter Mixture-of-Experts (16 experts, 2 active) model from Anthropic, featuring Grouped Query Attention (GQA) with 80 layers and 10,240 hidden dimensions. With a 1,000,000 token context window, it supports tools, vision, structured output, code, math, multilingual, reasoning. On standardized benchmarks, it achieves MMLU 91.5, HumanEval 75, GSM8K 97. The most cost-effective API deployment is via anthropic at $25.00/M output tokens. For self-hosted inference, B200 NVL (pair) delivers optimal throughput at $39858/month.

Provider pricing

1 provider · canonical: anthropic
Provider Input $/M Output $/M ▲Notes
anthropiccanonical$5.00$25.00cheapest input · cheapest output

Prices update via the nightly pricing cron + admin approvals at /admin/ingest-queue. The leaderboard's Input/Output cells show the canonical rate above; this table shows the full spread.

Recent changes

Loading…

Related models

5 suggestions

Picks: same family first, then same vendor within ±2× params, then top tag-overlap matches. Price shown is the cheapest Output $/M across providers — the row's page shows the canonical anchor.

Architecture Details

TypeMOE
Total Parameters450B
Active Parameters85B
Layers80
Hidden Dimension10,240
Attention Heads80
KV Heads10
Head Dimension128
Vocab Size200,000
Total Experts16
Active Experts2

Memory Requirements

BF16 Weights

900.0 GB

FP8 Weights

450.0 GB

INT4 Weights

225.0 GB

KV-Cache per Token204800 bytes
Activation Estimate4.00 GB

Fits on (single GPU) — most practical first

BEST FITB300288GBINT4B200 NVL (pair)360GBINT4

Fits on (multi-GPU with Tensor Parallelism)

RECOMMENDEDx2H200 SXM141GB eachINT4x2B200 SXM180GB eachINT4x2H100 NVL 94GB (per GPU pair)188GB eachINT4x2B100 SXM192GB eachINT4x2GB200 NVL72 (per GPU)192GB eachINT4x2GB300 NVL72 (per GPU)192GB eachINT4x2Instinct MI300X192GB eachINT4x2Groq LPU230GB eachINT4x2Instinct MI325X256GB eachINT4x3H100 NVL94GB eachINT4

Multi-GPU configurations use Tensor Parallelism (TP) to split model layers across GPUs. Requires NVLink or NVSwitch interconnect for optimal performance.

GPU Compatibility Matrix

Claude Opus 4.6 is compatible with 2% of GPU configurations across 41 GPUs at 3 precision levels.

BF16 (Full)
FP8 (Half)
INT4 (Quarter)
Blackwell(7 GPUs)
B200 NVL (pair)360GB
✗
✗
38%
B300288GB
✗
✗
22%
B100 SXM192GB
✗
✗
✗
GB200 NVL72 (per GPU)192GB
✗
✗
✗
Hopper(7 GPUs)
H100 NVL 94GB (per GPU pair)188GB
✗
✗
✗
H200 SXM141GB
✗
✗
✗
H2096GB
✗
✗
✗
GH20096GB
✗
✗
✗
Ada Lovelace(11 GPUs)
L40S48GB
✗
✗
✗
L4048GB
✗
✗
✗
RTX 6000 Ada48GB
✗
✗
✗
L2048GB
✗
✗
✗
Ampere(16 GPUs)
A100 80GB SXM80GB
✗
✗
✗
A100 80GB PCIe80GB
✗
✗
✗
A1664GB
✗
✗
✗
RTX A600048GB
✗
✗
✗
Legend:No fitVery tightTightModerateGoodExcellent

GPU Recommendations

B200 NVL (pair)optimal

BF16 · 4 GPUs · tensorrt-llm

88/100

score

Throughput

280.0 tok/s

Latency (ITL)

3.6ms

Est. TTFT

1ms

Cost/Month

$39858

Cost/M Tokens

$54.17

Use this config →
B200 SXMoptimal

BF16 · 8 GPUs · tensorrt-llm

83/100

score

Throughput

280.0 tok/s

Latency (ITL)

3.6ms

Est. TTFT

1ms

Cost/Month

$34088

Cost/M Tokens

$46.33

Use this config →
B100 SXMoptimal

BF16 · 8 GPUs · tensorrt-llm

83/100

score

Throughput

280.0 tok/s

Latency (ITL)

3.6ms

Est. TTFT

1ms

Cost/Month

$34164

Cost/M Tokens

$46.43

Use this config →

Deployment Options

API

API Deployment

anthropic

$25.00/M

output tokens

Self-Hosted

Single GPU

Requires multi-GPU setup (450 GB VRAM needed)

Scale

Multi-GPU

B200 NVL (pair) x4

280.0 tok/s

TP· $39858/mo

Open Full Calculator →

API Pricing Comparison

ProviderInput $/MOutput $/MBadges
anthropic$5.00$25.00
Cheapest

Cost Analysis

ProviderInput $/MOutput $/M~Monthly Cost
anthropicBest Value$5.00$25.00$150

Cost per 1,000 Requests

Short (500 tok)

$7.50

via anthropic

Medium (2K tok)

$30.00

via anthropic

Long (8K tok)

$90.00

via anthropic

Performance Estimates

Throughput by GPU

B200 NVL (pair)
280.0 tok/s
B200 SXM
280.0 tok/s
B100 SXM
280.0 tok/s

VRAM Breakdown (B200 NVL (pair), BF16)

Weights
Weights 225.0 GBKV-Cache 6.7 GBActivations 32.0 GBOverhead 11.3 GB

Quality Benchmarks

Top 10%
97th percentile across all models
MMLU
91.5
Top 10% (94th pctile)
HumanEval
75.0
Top 10% (91th pctile)
GSM8K
97.0
Top 10% (91th pctile)
MT-Bench
92.0
Bottom 25% (0th pctile)

Capabilities

Features

✓ Tool Use✓ Vision✓ Code✓ Math✓ Reasoning✓ Multilingual✓ Structured Output

Supported Frameworks

vllm

Supported Precisions

BF16 (default)

Where to Deploy Claude Opus 4.6

Similar Models

Claude Opus 4.5

400B params · moe

Quality: 90

from $25.00/M

Smaller contextCompare →

Claude Opus 4.7

500B params · moe

Quality: 90

from $25.00/M

Similar specsCompare →

Claude Opus 4.1

300B params · moe

Quality: 90

from $75.00/M

Smaller context, More expensiveCompare →

Claude Opus 4

200B params · dense

Quality: 90

from $75.00/M

Smaller context, More expensive, Smaller modelCompare →

Claude Sonnet 4.6

180B params · moe

Quality: 86

from $15.00/M

Cheaper, Smaller modelCompare →

Frequently Asked Questions

How much VRAM does Claude Opus 4.6 need for inference?

Claude Opus 4.6 requires approximately 900.0 GB of VRAM at BF16 precision, 450.0 GB at FP8, or 225.0 GB at INT4 quantization. Additional VRAM is needed for KV-cache (204800 bytes per token) and activations (~4.00 GB).

What is the best GPU for Claude Opus 4.6?

The top recommended GPU for Claude Opus 4.6 is the B200 NVL (pair) (x4) using BF16 precision. It achieves approximately 280.0 tokens/sec at an estimated cost of $39858/month ($54.17/M tokens). Score: 88/100.

How much does Claude Opus 4.6 inference cost?

Claude Opus 4.6 API inference starts from $5.00/M input tokens and $25.00/M output tokens. Self-hosted inference costs depend on your GPU configuration — use our ROI calculator for a detailed breakdown.