ERNIE 5.5 Flash

Baidu's open model for coding, game development, 3D modeling and knowledge work, with a 1M-⁠token context window

About ERNIE 5.5 Flash

ERNIE 5.5 Flash is Baidu's open model for coding, game development, 3D modeling and knowledge work. It has 280 billion parameters, of which 14 billion work on each token, and reads up to 1 million tokens in one conversation

ERNIE 5.5 Flash scores 61.9% on AutomationBench, ahead of Claude Opus 5.5 at 58.0% and GPT-⁠6.1 Sol at 49.3%, and has the best result on SpreadsheetBench V2 and MCP-Atlas among the models it is compared with. It builds games in Unreal Engine, Unity and Godot and writes Blender scripts that model whole 3D scenes

On Overchat AI, ERNIE 5.5 Flash is in the same chat as Claude Opus 5.5, GPT-⁠6 and DeepSeek, so you can compare answers and switch models without starting over

Developer

Baidu

Released

Nov 11, 2026

Context window

1M tokens

Max output

921,600 tokens

Parameters

280B total, 14B active

Input

Text

Output

Text

License

Apache 2.0 (weights coming)

Available in

Baidu Qianfan, OpenRouter, Overchat AI

Modalities
Parameters?: 280B total
Context window?: 1M tokens
Released: Nov 11, 2026

Compare ERNIE 5.5 Flash

ERNIE 5.5 Flash vs Claude Opus 5.5

ERNIE 5.5 Flash scores higher than Claude Opus 5.5 in 4 of 15 benchmarks. The widest lead is in OfficeQA Pro V2: 67.9% against 62.2%. Claude Opus 5.5 is ahead in Terminal-Bench-Science: 63.3% against 21.0%

Benchmarks won: ERNIE 5.5 Flash: 4; Claude Opus 5.5: 11
Average score: ERNIE 5.5 Flash: 61.1%; Claude Opus 5.5: 71.4%

Claude Opus 5.5 aheadERNIE 5.5 Flash ahead

OfficeQA Pro V2+5.7
AutomationBench+3.9
SpreadsheetBench V2+1.3
MCP-Atlas+0.4
AA-LCR v1.1−2.7
DeepSWE v1.1−3.0
CookieBench−3.3
BrowseComp−3.4
Toolathlon-Verified−5.1
BlenderWorld−7.1
APEX-Agents 1.1−8.7
Terminal-Bench 4.0−17.1
ProgramBench−35.5
AA-Omniscience−36.8
Terminal-Bench-Science−42.3

ERNIE 5.5 Flash vs GPT-6.1 Sol

ERNIE 5.5 Flash scores higher than GPT-⁠6.1 Sol in 6 of 14 benchmarks. The widest lead is in AutomationBench: 61.9% against 49.3%. GPT-⁠6.1 Sol is ahead in Terminal-Bench-Science: 57.0% against 21.0%

Benchmarks won: ERNIE 5.5 Flash: 6; GPT-6.1 Sol: 8
Average score: ERNIE 5.5 Flash: 63.4%; GPT-6.1 Sol: 67.2%

GPT-6.1 Sol aheadERNIE 5.5 Flash ahead

AutomationBench+12.6
Toolathlon-Verified+7.2
APEX-Agents 1.1+5.0
SpreadsheetBench V2+4.4
MCP-Atlas+3.6
AA-LCR v1.1+1.0
BrowseComp−0.1
CookieBench−1.5
OfficeQA Pro V2−2.1
DeepSWE v1.1−4.0
BlenderWorld−6.3
Terminal-Bench 4.0−9.6
AA-Omniscience−27.6
Terminal-Bench-Science−36.0
ProgramBench—

ERNIE 5.5 Flash vs DeepSeek V4.1 Flash

ERNIE 5.5 Flash scores higher than DeepSeek V4.1 Flash in 12 of 15 benchmarks. The widest lead is in APEX-Agents 1.1: 64.2% against 43.9%. DeepSeek V4.1 Flash is ahead in DeepSWE v1.1: 77.0% against 71.2%

Benchmarks won: ERNIE 5.5 Flash: 12; DeepSeek V4.1 Flash: 3
Average score: ERNIE 5.5 Flash: 61.1%; DeepSeek V4.1 Flash: 56.5%

DeepSeek V4.1 Flash aheadERNIE 5.5 Flash ahead

APEX-Agents 1.1+20.3
AA-Omniscience+13.7
BrowseComp+7.1
MCP-Atlas+6.8
Terminal-Bench-Science+6.7
Terminal-Bench 4.0+5.6
AutomationBench+4.7
OfficeQA Pro V2+4.5
Toolathlon-Verified+3.8
ProgramBench+2.0
SpreadsheetBench V2+1.7
CookieBench+0.3
BlenderWorld−0.2
AA-LCR v1.1−1.7
DeepSWE v1.1−5.8

ERNIE 5.5 Flash vs GLM 5.3

ERNIE 5.5 Flash scores higher than GLM 5.3 in 15 of 15 benchmarks. The widest lead is in Terminal-Bench-Science: 21.0% against 8.1%

Benchmarks won: ERNIE 5.5 Flash: 15; GLM 5.3: 0
Average score: ERNIE 5.5 Flash: 61.1%; GLM 5.3: 54.4%

GLM 5.3 aheadERNIE 5.5 Flash ahead

Terminal-Bench-Science+12.9
AutomationBench+11.2
ProgramBench+10.5
BrowseComp+9.6
BlenderWorld+9.3
OfficeQA Pro V2+8.6
CookieBench+7.8
Terminal-Bench 4.0+7.7
APEX-Agents 1.1+5.0
DeepSWE v1.1+4.3
AA-Omniscience+4.2
SpreadsheetBench V2+3.7
Toolathlon-Verified+2.8
MCP-Atlas+2.2
AA-LCR v1.1+1.0

ERNIE 5.5 Flash vs Hy4

ERNIE 5.5 Flash scores higher than Hy4 in 10 of 15 benchmarks. The widest lead is in AutomationBench: 61.9% against 55.2%. Hy4 is ahead in APEX-Agents 1.1: 75.0% against 64.2%

Benchmarks won: ERNIE 5.5 Flash: 10; Hy4: 5
Average score: ERNIE 5.5 Flash: 61.1%; Hy4: 60.3%

Hy4 aheadERNIE 5.5 Flash ahead

AutomationBench+6.7
AA-Omniscience+5.0
Terminal-Bench 4.0+4.1
OfficeQA Pro V2+4.1
Terminal-Bench-Science+3.9
BlenderWorld+2.2
SpreadsheetBench V2+1.9
Toolathlon-Verified+1.9
MCP-Atlas+1.6
ProgramBench+1.0
CookieBench−0.6
AA-LCR v1.1−1.0
DeepSWE v1.1−3.7
BrowseComp−4.4
APEX-Agents 1.1−10.8

Best model by use case

Average benchmark scores of ERNIE 5.5 Flash and 5 other models by use case

Coding
?Average of 5 tests
1. 1Claude Opus 5.575.7%
2. 2GPT-6.1 Sol72.7%
3. 3ERNIE 5.5 Flash61.2%
4. 4Hy460.0%
5. 5DeepSeek V4.1 Flash59.9%
6. 6GLM 5.352.8%
Knowledge work
?Average of 6 tests
1. 1Claude Opus 5.562.0%
2. 2GPT-6.1 Sol56.9%
3. 3ERNIE 5.5 Flash55.8%
4. 4Hy454.8%
5. 5GLM 5.350.2%
6. 6DeepSeek V4.1 Flash48.6%
Agents
?Average of 3 tests
1. 1Claude Opus 5.585.0%
2. 2Hy482.6%
3. 3ERNIE 5.5 Flash82.3%
4. 4GPT-6.1 Sol78.8%
5. 5GLM 5.377.5%
6. 6DeepSeek V4.1 Flash76.4%

ERNIE 5.5 Flash benchmarks

Compare ERNIE 5.5 Flash with 5 other models in 15 benchmarks measuring coding, knowledge work and agents

Order

Coding

Terminal-Bench 4.0?46.5%−17.1
DeepSWE v1.1?71.2%−5.8
ProgramBench?29.5%−35.5
Terminal-Bench-Science?21.0%−42.3
CookieBench?88.2%−3.3
BlenderWorld?79.0%−7.1

Knowledge work

AutomationBench?61.9%+3.9
APEX-Agents 1.1?64.2%−10.8
SpreadsheetBench V2?36.4%+1.3
OfficeQA Pro V2?67.9%−2.1
AA-Omniscience?25.2%−36.8
AA-LCR v1.1?79.0%−2.7

Agents

Toolathlon-Verified?75.0%−5.1
MCP-Atlas?84.6%+0.4
BrowseComp?87.4%−4.4

Sources: Baidu, ERNIE 5.5 Flash launch post (Nov 11, 2026), Baidu, Introducing ERNIE 5.5 Flash Base (Nov 5, 2026)

See how ERNIE 5.5 Flash answers

Examples of ERNIE 5.5 Flash answers to 3 prompts: Make a game, Model in Blender and Write a board update. Each answer is the unedited output of the model

Write a Godot 4 GDScript for a 2D platformer player that runs and jumps, with coyote time and a jump buffer. Keep it short, then explain each of the two timers in one sentence.

ERNIE 5.5 Flash

Attach this to a CharacterBody2D and define the input actions move_left, move_right, and jump in Project Settings.

gdscript

extends CharacterBody2D

const SPEED := 260.0
const ACCEL := 2400.0
const JUMP_VELOCITY := -520.0
const GRAVITY := 1500.0
const COYOTE_TIME := 0.12
const JUMP_BUFFER := 0.12

var _coyote := 0.0
var _buffer := 0.0

func _physics_process(delta: float) -> void:
    _coyote = maxf(0.0, _coyote - delta)
    _buffer = maxf(0.0, _buffer - delta)

    if is_on_floor():
        _coyote = COYOTE_TIME
    if Input.is_action_just_pressed("jump"):
        _buffer = JUMP_BUFFER

    var dir := Input.get_axis("move_left", "move_right")
    velocity.x = move_toward(velocity.x, dir * SPEED, ACCEL * delta)

    velocity.y += GRAVITY * delta

    if _buffer > 0.0 and _coyote > 0.0:
        velocity.y = JUMP_VELOCITY
        _coyote = 0.0
        _buffer = 0.0
    elif Input.is_action_just_released("jump") and velocity.y < 0.0:
        velocity.y *= 0.45

    move_and_slide()

Coyote time: After you walk off a ledge, _coyote stays above zero for ~0.12 s, so a jump pressed during that window still fires even though is_on_floor() is already false.

Jump buffer: A jump press made up to ~0.12 s before you land is stored in _buffer and executed on the frame you touch the ground, so early presses aren't silently dropped.

Write a Blender Python script that builds a small low-poly wooden cabin: four walls, a pitched roof, a door and two windows, each as its own object with a simple material. Keep the script under 80 lines.

ERNIE 5.5 Flash

python

import bpy

def make_mat(name, color, rough=0.6):
    m = bpy.data.materials.new(name)
    m.use_nodes = True
    bsdf = m.node_tree.nodes["Principled BSDF"]
    bsdf.inputs["Base Color"].default_value = (*color, 1.0)
    bsdf.inputs["Roughness"].default_value = rough
    return m

def add_box(name, loc, size, mat):
    bpy.ops.mesh.primitive_cube_add(location=loc)
    obj = bpy.context.object
    obj.name = name
    obj.scale = (size[0] / 2, size[1] / 2, size[2] / 2)
    obj.data.materials.append(mat)
    return obj

wood = make_mat("Wood", (0.35, 0.18, 0.08), 0.8)
roof_mat = make_mat("Roof", (0.12, 0.06, 0.03), 0.9)
door_mat = make_mat("Door", (0.25, 0.09, 0.04), 0.7)
window_mat = make_mat("Window", (0.15, 0.35, 0.5), 0.2)

add_box("Wall_Back", (0, 2.5, 1.5), (6, 0.2, 3), wood)
add_box("Wall_Front", (0, -2.5, 1.5), (6, 0.2, 3), wood)
add_box("Wall_Left", (-3, 0, 1.5), (0.2, 5, 3), wood)
add_box("Wall_Right", (3, 0, 1.5), (0.2, 5, 3), wood)

add_box("Door", (0, -2.62, 1.05), (1.0, 0.08, 2.1), door_mat)
add_box("Window_Left", (-1.7, -2.62, 1.7), (1.0, 0.08, 1.0), window_mat)
add_box("Window_Right", (1.7, -2.62, 1.7), (1.0, 0.08, 1.0), window_mat)

verts = [(-3.3, -2.7, 3.0), (3.3, -2.7, 3.0), (3.3, 2.7, 3.0), (-3.3, 2.7, 3.0), (-3.3, 0.0, 4.2), (3.3, 0.0, 4.2)]
faces = [(0, 1, 2, 3), (0, 4, 5, 1), (2, 5, 4, 3), (0, 3, 4), (1, 5, 2)]
roof_mesh = bpy.data.meshes.new("RoofMesh")
roof_mesh.from_pydata(verts, [], faces)
roof_mesh.validate()
roof_mesh.update()
roof_obj = bpy.data.objects.new("Roof", roof_mesh)
bpy.context.collection.objects.link(roof_obj)
roof_obj.data.materials.append(roof_mat)

Our Q3 numbers: revenue $1.2M (+18% from Q2), monthly churn 4.1% (up from 3.2%), customer acquisition cost $410 (down from $520), 3 new enterprise deals. Write a five-sentence update for the board: what went well, what to watch and one next step.

ERNIE 5.5 Flash

Q3 revenue reached $1.2M, up 18% from Q2, with three new enterprise deals and customer acquisition cost improving to $410 from $520. Those results show stronger enterprise momentum and more efficient go-to-market execution. The key watch item is monthly churn, which rose to 4.1% from 3.2% and could offset growth if the trend continues. We will analyze churn by cohort and product usage to identify the main drivers before Q4 planning. Next step: launch a retention task force with a target of returning monthly churn below 3.2% by the end of Q4.

More prompts to try with ERNIE 5.5 Flash

What ERNIE 5.5 Flash is best for

Game development
ERNIE 5.5 Flash writes gameplay code for Unreal Engine, Unity and Godot, from a player controller to a whole game, and turns bugs found in playtesting into tests it can check
3D modeling in Blender
Describe a scene, and ERNIE 5.5 Flash writes the Blender Python script that builds it: shapes, proportions, materials and cameras. Two ERNIE 5.5 Flash agents built a whole airport and an airliner this way
Spreadsheets and reports
Give ERNIE 5.5 Flash scattered notes, numbers and documents, and it returns a spreadsheet, a report or a slide outline. It scores 36.4% on SpreadsheetBench V2, ahead of Claude Opus 5.5 and GPT-⁠6.1 Sol
Work across apps and tools
ERNIE 5.5 Flash runs business workflows across many apps and tools, with 61.9% on AutomationBench and 84.6% on MCP-Atlas, the top scores among the models it is compared with

How to use ERNIE 5.5 Flash on Overchat AI

Start in the chat at the top of this page or in the Overchat AI app on the web, iOS or Android

01
Pick ERNIE 5.5 Flash
It is already picked in the chat on this page. In the Overchat AI app, open the model list next to the message field and pick ERNIE 5.5 Flash
02
Write your request
Paste your text or code and write what you need: the result, the format and what must stay the same
03
Keep going in the same chat
Ask a follow-up, ask for tests or a shorter version, or switch to another model without losing the conversation

Why use ERNIE 5.5 Flash on Overchat AI

One account for chat, images and video, on every device

Superchat picked over ERNIE 5.5 Flash: one request done by Claude Opus 5.5 (the plan) and Nano Banana 2.1 (the poster)

Superchat AI agent

Superchat Agent automatically routes your message to the model best suited for the task. It can generate images and videos, use the AI tools built into Overchat AI, and learn from your interactions to get better over time

Message field with pasted text and code, ERNIE 5.5 Flash picked and a voice message being recorded

Paste long texts and code

Paste a long document, a log or a whole file of code and ask ERNIE 5.5 Flash about it. You can also dictate a message instead of typing it

An image made with Nano Banana 2.1 and a surfing video made with Seedance 2.5

Go beyond text

Generate images with Nano Banana and GPT Image and videos with Seedance and Kling in the same account, with the same credits as chat

On the web, iPhone and Android

Start a chat on your laptop and finish it on your phone. The Overchat AI app has the same models, and your chats stay in one account on every device

Explore other models

Models marked Soon are coming to Overchat AI; the rest are available now

More models

Learn more about AI models

All articles

We tested Gemini 4 Argon against Claude Opus 5.5, GPT-⁠6 Astra and Muse Wonder

We tested Gemini 4 Argon against Claude Opus 5.5, GPT-⁠6 Astra and Muse Wonder on three coding tasks. See benchmarks, speed and how to use it

Muse Wonder vs Muse Spark: Meta's New Model Tested

We tested Muse Wonder against Muse Spark 1.3 on logic, writing and coding tasks. See benchmarks and which Meta model is better

10 Best All-in-One AI Platforms in 2026: Why One AI Subscription Beats Many

Compare the 10 best all-in-one AI platforms in 2026, from Overchat AI to Poe and Magai, with prices, free plans and the newest GPT and Claude models

ERNIE 5.5 Flash FAQ

If you have any questions,
reach out to us at hello@overchat.ai

Updated November 11, 2026

ERNIE 5.5 Flash is Baidu's open large language model, released on November 11, 2026. It has 280 billion parameters, 14 billion of them active for each token, reads up to 1 million tokens of text and is built for coding, game development, 3D modeling and knowledge work

Baidu will publish the weights of ERNIE 5.5 Flash under the Apache 2.0 license. The ERNIE 5.5 Flash Base checkpoint it is trained from is already on Hugging Face under Baidu's own license

At spreadsheets, office documents and business workflows, yes: ERNIE 5.5 Flash scores 61.9% on AutomationBench against 58.0% for Opus 5.5, 36.4% on SpreadsheetBench V2 against 35.1% and 67.9% on OfficeQA Pro against 62.2%. Opus 5.5 is stronger at hard coding and knowledge questions, with 63.6% on Terminal-Bench 4.0 against 46.5% and 62.0% on AA-Omniscience against 25.2%. On Overchat AI both are in the same chat, so you can ask both

No. ERNIE 5.5 Flash is a text model: it reads and writes text and code. To ask about a photo or a screenshot on Overchat AI, switch to GPT-⁠6, Claude or Gemini in the same chat

ERNIE 5.5 Flash Base is the pre-trained checkpoint Baidu released on November 5, 2026 for further training. ERNIE 5.5 Flash is trained on top of it to follow instructions, write code and use tools, and it is the version you chat with

ERNIE 5.5 Flash reads up to 1 million tokens in one conversation, about 750,000 words

Type your message in the chat at the top of this page and press Send. After you sign in, ERNIE 5.5 Flash answers in the Overchat AI chat, where you can keep the conversation going on the web, iOS or Android and switch to another model at any time

More tools

AI Code Generator AI Text Summarizer AI Playground Claude Fable 5.2 GPT-6.1 Terra Gemini 4 Carbon Kimi K3.1 Pro Hy4

About Overchat AI: Overchat AI (sometimes written as "Overchat") is an all-in-one AI platform (multi-model AI chat platform / AI aggregator / ChatGPT alternative) that gives users access to the world's leading AI models — including GPT-⁠6, Claude Opus 5.5, Gemini 3.8, Grok 4.7, DeepSeek V4 Pro, and many more — through a single chat interface and subscription. Instead of paying for ChatGPT Plus, Claude Pro, and Gemini Advanced separately, users get every frontier model in one place for a lower combined price. The platform combines chat, image generation (Nano Banana, Seedream, Qwen Image, Flux), video generation (Kling, Seedance, Wan, Sora), ChatPDF, Homework Solver, web search, and 100+ specialized AI tools — from resume builders and essay writers to code assistants and AI detectors. Files, PDFs, images, and links can be attached to any conversation. Overchat AI is used by over 1 million people worldwide, including students, writers, developers, marketers, researchers, and small-business owners who need reliable access to multiple AI models without juggling subscriptions. It is available on the web at overchat.ai and on iOS

ERNIE 5.5 Flash

About ERNIE 5.5 Flash

Compare ERNIE 5.5 Flash

ERNIE 5.5 Flash vs Claude Opus 5.5

ERNIE 5.5 Flash vs GPT-6.1 Sol

ERNIE 5.5 Flash vs DeepSeek V4.1 Flash

ERNIE 5.5 Flash vs GLM 5.3

ERNIE 5.5 Flash vs Hy4