Back to LeaderboardCompare Hub
Model profile
Baidu

Baidu: ERNIE 4.5 21B A3B Thinking

Baidu: ERNIE 4.5 21B A3B Thinking is a budget-priced text-first model from Baidu with partial runtime data, extended context posture, and the clearest fit around long-context research.

Best for: Long-context researchN/A latencyExtended contextBudget pricing
Coding models
Intelligence
N/A

Benchmark blend

Coding
N/A

Dev workflow signal

Context
131K Tokens

Extended

Input Price
$0.07

Budget tier

Decision snapshot
81

Baidu: ERNIE 4.5 21B A3B Thinking currently reads as a budget text-first option with extended context and a partially published runtime profile.

Overall profile
Flagship profile
Best for
Long-context research
Latency tier
N/A
Price tier
Budget
Source coverage
OpenRouter
Open compare hubSee leaderboard

Decision Strip

Decision rail before the raw tables

Core buy-side signals stay in one pass. The rest of the page expands only after intelligence, speed, context, and price are clear.

Intelligence
N/A
N/A

General reasoning and benchmark headroom.

Unavailable
Speed
N/A
N/A

Latency data is partial.

Unavailable
Context
131K Tokens
76

How much prompt and task state can stay in view.

Competitive
Price
$0.07
86

$0.28 output / 1M

Efficient

Editorial Profile

Baidu: ERNIE 4.5 21B A3B Thinking in one narrative

Positioning, tradeoffs, and fit are consolidated into one read instead of repeating the same story across separate cards.

Flagship profileCoding score N/AMath score N/A

ERNIE-4.5-21B-A3B-Thinking is Baidu's upgraded lightweight MoE model, refined to boost reasoning depth and quality for top-tier performance in logical puzzles, math, science, coding, text generation, and expert-level academic benchmarks.

Identity

Baidu text-first profile

Positioning

Long-context research with extended context and partially published runtime.

Cost posture

Efficient spend profile. More comfortable for sustained prompt volume if the capability fit is right.

Strengths

  • Large context headroom supports repo-wide prompts and long research sessions.

Tradeoffs

  • Budget-friendly input pricing is a strength, but raw capability may vary by workload.

  • Latency data is incomplete, so interactive responsiveness is harder to rank confidently.

  • Current metadata points to a text-first profile rather than a broad multimodal one.

Best fit

  • Long-context summarization, repo analysis, and policy or document review.

Explore Next

Similar profiles worth opening next

Baidu

ERNIE 5.0 Thinking Preview

Intelligence
29.1
Context
N/A
Input Price
$0.00
Open profile
Baidu

Baidu: ERNIE 4.5 300B A47B

Intelligence
15.0
Context
123K Tokens
Input Price
$0.28
Open profile
Baidu

Baidu: ERNIE 4.5 21B A3B

Intelligence
N/A
Context
120K Tokens
Input Price
$0.07
Open profile

Benchmarks

Grouped by job-to-be-done

Only benchmark categories with actual signal are shown. Secondary values stay as simple definitions instead of nested micro-cards.

No benchmark data is available for this model yet.

Specs & Pricing

Technical snapshot and cost posture

Specs stay neutral, pricing gets emphasis through values rather than extra containers. Raw provider internals remain in metadata at the end.

Technical snapshot
Context Window
131K Tokens
Vision
Text-first
Modalities
text
Tokenizer
Other
Max Completion
65536
Moderation
No
Supported Parameters
frequency_penaltyinclude_reasoningmax_tokenspresence_penaltyreasoningrepetition_penaltyseedstoptemperaturetop_ktop_p
Input Modalities
text
Output Modalities
text
Price architecture
Input
per 1M input tokens
$0.07
Output
per 1M output tokens
$0.28
Blended
AA 3:1 mix
N/A

This model is relatively efficient on price. It is the easier fit when sustained prompt volume matters.

Metadata

Raw source tables at the end

Verification details remain available, but the page no longer forces them ahead of the editorial read.