vs
0:2
overall capability wins (1 ties across 3 scored). Claude Haiku 4.5 leads overall.
- Arena0:1
- Pricing0:3
- Speed0:2
- Specs0:1
Verdict
Claude Haiku 4.5 ranks higher on LMArena (+134 Elo); Claude Haiku 4.5 allows ~15.6x more max output (64K vs 4K); Claude Haiku 4.5 is ~3.0x cheaper on a blended token basis than Claude 3 Sonnet.
- Claude Haiku 4.5 ranks higher on LMArena (+134 Elo)
- Claude Haiku 4.5 allows ~15.6x more max output (64K vs 4K)
- Claude Haiku 4.5 is ~3.0x cheaper on a blended token basis than Claude 3 Sonnet
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Claude 3 Sonnet
- $6.75
- Claude Haiku 4.5
- $2.25
Claude Haiku 4.5 is about 3.0x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
Anthropic
Anthropic
Family
Claude Sonnet / Haiku
Claude Sonnet / Haiku
License
Proprietary
Proprietary
Open weights
No
No
Release
Mar 4, 2024
Oct 15, 2025
Knowledge cutoff
-
-
API / provider
Anthropic
Anthropic
Modalities
text, image → text
text, image → text
Specs
Context window
200K
200K
Max output
4K
64K
Parameters
—
—
Pricing
Input $/1M
$3
$1
Output $/1M
$15
$5
Blended $/1M (3∶1)
$6
$2
Speed
tok/s
65
150
TTFT (s)
0.5
0.2
Reasoning
MMLU-Pro
57.9
—
GPQA Diamond
40.0
—
Humanity's Last Exam
3.6
—
AIME 2025
—
—
MATH-500
41.4
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
—
—
AA-LCR v1.1
—
—
CritPt
—
—
MMMU-Pro
—
—
IFBench
—
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
73.3
SWE-bench Pro
—
39.5
SWE-bench Multilingual
—
64.7
LiveCodeBench
17.5
—
Terminal-Bench 2.1
—
—
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
42.7
—
SciCode
22.9
—
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
1,329
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
1,281
1,415
| Metric | Claude 3 Sonnet | Claude Haiku 4.5 | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Anthropic | Anthropic | — |
| Family | Claude Sonnet / Haiku | Claude Sonnet / Haiku | — |
| License | Proprietary | Proprietary | — |
| Open weights | No | No | — |
| Release | Mar 4, 2024 | Oct 15, 2025 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Anthropic | Anthropic | — |
| Modalities | text, image → text | text, image → text | — |
| Specs | |||
| Context window | 200K | 200K | tie |
| Max output | 4K | 64K | B +60K |
| Parameters | — | — | — |
| Pricing | |||
| Input $/1M | $3 | $1 | B +$2 |
| Output $/1M | $15 | $5 | B +$10 |
| Blended $/1M (3∶1) | $6 | $2 | B +$4 |
| Speed | |||
| tok/s | 65 | 150 | B +85 |
| TTFT (s) | 0.5 | 0.2 | B +0.30 |
| Reasoning | |||
| MMLU-Pro | 57.9 | — | — |
| GPQA Diamond | 40.0 | — | — |
| Humanity's Last Exam | 3.6 | — | — |
| AIME 2025 | — | — | — |
| MATH-500 | 41.4 | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | — | — | — |
| AA-LCR v1.1 | — | — | — |
| CritPt | — | — | — |
| MMMU-Pro | — | — | — |
| IFBench | — | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | 73.3 | — |
| SWE-bench Pro | — | 39.5 | — |
| SWE-bench Multilingual | — | 64.7 | — |
| LiveCodeBench | 17.5 | — | — |
| Terminal-Bench 2.1 | — | — | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | 42.7 | — | — |
| SciCode | 22.9 | — | — |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | 1,329 | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | 1,281 | 1,415 | B +134 Elo |
Claude 3 Sonnet
- Context
- 200K / 4K out
- Parameters
- —
- Price
- $3 / $15
- Speed
- 65 tok/s · 0.5s TTFT
- Modalities
- text, image → text
- License
- Proprietary
Balanced Claude 3 mid-tier between Haiku and Opus — superseded by 3.5 / 4.x Sonnet.
Claude Haiku 4.5
- Context
- 200K / 64K out
- Parameters
- —
- Price
- $1 / $5
- Speed
- 150 tok/s · 0.2s TTFT
- Modalities
- text, image → text
- License
- Proprietary
Fast, affordable Claude tier for classification, routing, and lightweight assistants.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.