Current model · Baidu
ERNIE 5.5 Flash Preview benchmarks, pricing and specs
ERNIE 5.5 Flash Preview is a Baidu model released on November 11, 2026 that takes text input and produces text. It costs $0.28 per 1M input tokens and $1.13 per 1M output tokens and accepts up to 1,000,000 tokens of context.
- Per 1M input tokens
- $0.28Listed today
- Per 1M output tokens
- $1.13Listed today
- Context window
- 1M#29 of 67 current models
Data verified November 11, 2026 · Dataset 2026-11-11
Specs
Key facts
- Provider
- Baidu
- Model family
- ERNIE 5.5
- Release date
- November 11, 2026
- Status
- Current
- Access
- API only
- API model id
- ernie-5.5-flash-preview
- Context window
- 1,000,000 tokens
- Max output
- 1,000,000 tokens
- Input
- text
- Output
- text
- Task families
- text
- Declared capabilities
- reasoning, coding, agentic coding
Charts
Against every current text model
Where ERNIE 5.5 Flash Preview lands among the current models of its task, on every independent benchmark, on price and on context.
Where it stands
11 of 57 list a lower price per 1M input tokens. Each bar is the share of current models it beats.
Price and context
Price per 1M input tokens
#12of 57
$0.28 · cheaper than 80%
Price per 1M output tokens
#12of 57
$1.13 · cheaper than 80%
Context window
#29of 67
1M tokens · longer than 24%
Benchmarks
Independent benchmark results
Scores come from independent evaluators that ran the model themselves, pinned to a dated snapshot. Rank is among current models in the same snapshot. Baidu's own launch figures are in the next section, because they are not measured the same way.
No independent benchmark has published a result for ERNIE 5.5 Flash Preview yet. It is still compared below on price, context and capabilities.
Provider results
Reported by Baidu
Baidu's own results for ERNIE 5.5 Flash Preview, beside the rivals it chose, as its table lists them. Labs pick the benchmarks and run them their own way, so compare within one table only. Internal benchmarks and other evaluators' indexes are left out.
ERNIE 5.5 Flash Preview launch table, November 11, 2026Baidu · 2026-11-11
| Benchmark | ERNIE 5.5 Flash Preview | Hy4 | DeepSeek-V4.1-Flash | GLM-5.3-Flash | GLM-5.3 | GPT-6.1 Sol | Claude Opus 5.5 |
|---|---|---|---|---|---|---|---|
| Terminal-Bench 4.0 | 46.5% | 42.4% | 40.9% | 37.9% | 38.8% | Not listed | 63.6% |
| DeepSWE v1.1 | 71.2% | 74.9% | 77.0% | 63.4% | 66.9% | 75.2% | 74.2% |
| ProgramBenchalmost solved | 29.5% | 28.5% | 27.5% | 15.0% | 19.0% | Not listed | 65.0% |
| NL2Repo | 64.8% | 65.8% | 62.7% | 57.5% | 58.8% | Not listed | Not listed |
| Terminal-Bench-Science 0.1 | 21.0% | 17.1% | 14.3% | 5.7% | 8.1% | 57.0% | 63.3% |
| CookieBench | 88.2% | 88.8% | 87.9% | 76.9% | 80.4% | 89.7% | 91.5% |
| BlenderWorldtext-only subset | 79.0% | 76.8% | 79.2% | 71.4% | 69.7% | 85.3% | 86.1% |
| AutomationBench v1.0.6open-sourced subset | 61.9% | 55.2% | 57.2% | 48.7% | 50.7% | 49.3% | 58.0% |
| APEX-Agents v1.1 | 64.2% | 75.0% | 43.9% | 52.9% | 59.2% | 59.2% | 72.9% |
| DRACO | 84.9% | 90.7% | 83.6% | 78.4% | 82.3% | 79.4% | 87.9% |
| SpreadsheetBench 2 | 36.4% | 34.5% | 34.7% | 31.5% | 32.7% | 32.0% | 35.1% |
| OfficeQA Pro v2 | 67.9% | 63.8% | 63.4% | 64.4% | 59.3% | 70.0% | 62.2% |
| Agents' Last Examcommand-line tasks | 29.0% | 32.4% | 35.2% | 28.6% | 22.9% | 29.5% | 32.4% |
| SkillsBench v1.1text-only subset | 73.5% | 77.4% | 67.5% | 65.4% | 62.0% | 72.6% | 71.8% |
| Toolathlon-Verified | 75.0% | 73.1% | 71.2% | 76.8% | 72.2% | 67.8% | 80.1% |
| MCP Atlaspublic set | 84.6% | 83.0% | 77.8% | 78.8% | 82.4% | 81.0% | 84.2% |
| BrowseComp | 87.4% | 91.8% | 80.3% | 74.6% | 77.8% | 87.5% | 90.8% |
| Humanity's Last Examwithout tools, text-only | 46.4% | 46.7% | 44.0% | Not listed | 42.0% | Not listed | Not listed |
Baidu's appendix to the ERNIE 5.5 Flash Preview announcement, every model at its highest reasoning setting. Where Baidu lists a lab's reported score beside its own run, the score Baidu measured is shown. DeepSeek-V4.1-Flash is the 1017 build. Left out here: AA-Omniscience-Public and AA-LCR (Artificial Analysis benchmarks), GPT-6.1 Sol on Terminal-Bench 4.0 (cited from Artificial Analysis), and the full-set Humanity's Last Exam scores in a row Baidu otherwise ran on the text-only subset.
Fit
Strengths, cautions and best uses
- Strengths
- Baidu's first post-trained model on its ERNIE 5.5 Flash base: a 280B-parameter mixture-of-experts with 14B active per token and a 1M-token context, built for long agent runs in coding, 3D modeling and office work
- In Baidu's own table it scores highest of all seven models on AutomationBench (61.9%), SpreadsheetBench V2 (36.4%) and MCP-Atlas (84.6%), Claude Opus 5.5 and GPT-6.1 Sol included
- $0.28 input, $1.13 output and $0.028 cached input per 1M tokens on Qianfan, the same input and output rates Qianfan charges for DeepSeek-V4.1-Flash
- Cautions
- A preview, served only through Baidu Qianfan: Baidu says the weights will be open-sourced under Apache 2.0 soon, but they are not out yet
- Text only: it cannot read images, and in Baidu tests that needed vision a tool running Claude Opus 5.5 described the screen for it
- In Baidu's own table it trails Hy4 on APEX-Agents 1.1 (64.2% against 75.0%) and DeepSeek-V4.1-Flash on DeepSWE v1.1 (71.2% against 77.0%), and stays far behind GPT-6.1 Sol and Claude Opus 5.5 on Terminal-Bench-Science 0.1 (21.0% against 57.0% and 63.3%)
- Epoch AI and Arena have not scored it yet
- Best for
- Low-cost agents for spreadsheets, office documents and tool calls over MCP, where text input is enough
- Long-horizon coding and game or 3D scene building inside an agent harness, at a Flash-tier price
From benchmark to production
Building on ERNIE 5.5 Flash Preview?
Public scores rarely hold on your own data. We test it against the alternatives on your real tasks, then build the agent or pipeline around the model that wins.
AI Model Cost AuditLaunch price $100, regular price $150What the audit includes
Alternatives
Alternatives to ERNIE 5.5 Flash Preview
The closest current text models by price. The cheapest model that scores as well or better is listed first; failing that, the cheapest one within 3% of the benchmark's scale.
On score and price
ERNIE 5.5 Flash Preview beside the 4 alternatives listed below, on Arena Text and on blended price per 1M tokens.
Arena Text
Measured November 9, 2026, higher is better
- ERNIE 5.5 Flash Preview
- Not scored
- DeepSeek-V4.1-Flash
- 1,475
- Gemini 3.5 Flash-Lite
- 1,455
- Inkling 2 Small
- 1,412
- LongCat-2.5
- Not scored
Blended price per 1M tokens
Listed today, lower is better
- ERNIE 5.5 Flash Preview
- $0.493
- DeepSeek-V4.1-Flash
- $0.525
- Gemini 3.5 Flash-Lite
- $0.85
- Inkling 2 Small
- $0.475
- LongCat-2.5
- $0.525
DeepSeek-V4.1-Flash
- 7% more per 1M input tokens, $0.30 vs $0.28
- From DeepSeek
Gemini 3.5 Flash-Lite
- 7% more per 1M input tokens, $0.30 vs $0.28
- From Google
Inkling 2 Small
- 7% more per 1M input tokens, $0.30 vs $0.28
- 262K context window vs 1M
- From Thinking Machines Lab
LongCat-2.5
- 7% more per 1M input tokens, $0.30 vs $0.28
- From Meituan
Further reading
More on ERNIE 5.5 Flash Preview
Compare ERNIE 5.5 Flash Preview with
- ERNIE 5.5 Flash Preview vs DeepSeek-V4.1-Flash
- ERNIE 5.5 Flash Preview vs Gemini 3.5 Flash-Lite
- ERNIE 5.5 Flash Preview vs GLM-5.3
- ERNIE 5.5 Flash Preview vs GLM-5.3-Flash
- ERNIE 5.5 Flash Preview vs Hy4
- ERNIE 5.5 Flash Preview vs Inkling 2 Small
- ERNIE 5.5 Flash Preview vs LongCat-2.5
FAQ
ERNIE 5.5 Flash Preview FAQ
How much does ERNIE 5.5 Flash Preview cost?
ERNIE 5.5 Flash Preview costs $0.28 per 1M input tokens and $1.13 per 1M output tokens at the standard tier. It is the #12 cheapest of 57 current text models priced per 1M input tokens. Source: Baidu, verified November 11, 2026.
What is the context window of ERNIE 5.5 Flash Preview?
ERNIE 5.5 Flash Preview accepts up to 1,000,000 tokens of context. It can return up to 1,000,000 tokens in one response. That ranks #29 of 67 current text models by context length.
What benchmark results has Baidu published for ERNIE 5.5 Flash Preview?
In the ERNIE 5.5 Flash Preview launch table (November 11, 2026), Baidu reports 46.5% on Terminal-Bench 4.0, 71.2% on DeepSWE v1.1, 29.5% on ProgramBench (almost solved) and 64.8% on NL2Repo. These are Baidu's own figures, beside the rivals it chose, not independent measurements.
Is ERNIE 5.5 Flash Preview open source or API only?
ERNIE 5.5 Flash Preview is available through Baidu's API only; its weights are not published.
What are the best alternatives to ERNIE 5.5 Flash Preview?
The closest current alternatives by price and measured performance are DeepSeek-V4.1-Flash (7% more per 1M input tokens, $0.30 vs $0.28), Gemini 3.5 Flash-Lite (7% more per 1M input tokens, $0.30 vs $0.28), Inkling 2 Small (7% more per 1M input tokens, $0.30 vs $0.28) and LongCat-2.5 (7% more per 1M input tokens, $0.30 vs $0.28).
Sources
Sources
- Introducing ERNIE 5.5 Flash Preview · Baidu · official · verified November 11, 2026
- Qianfan model list (international) · Baidu · official · verified November 11, 2026
- Qianfan model inference pricing (international) · Baidu · official · verified November 11, 2026
All 273 sources behind the benchmarks, with their versions, licenses and notes, are in the source catalog.
Cite this page
Van Data Team. "ERNIE 5.5 Flash Preview Benchmarks, Pricing and Specs." AI Model Benchmark Dataset 2026-11-11. https://vandatateam.com/ai-model-benchmarks/ernie-5-5-flash-preview
Reuse the data: JSON · CSV · or explore every model in the AI model benchmark comparison.
Production evaluation
Deciding whether ERNIE 5.5 Flash Preview fits your workload?
Public benchmarks narrow the shortlist. We test the finalists on your own tasks, data and cost limits, then build the routing and review controls to run the winner in production.
- Evaluation cases and acceptance thresholds tied to real work
- Model routing, fallbacks and cost controls