Hardware FixRecommendedDevice not working? Your driver may be the problemCheck updates for common hardware issues.Fix DriversOctober DealsAmazon USOctober deal check: compare before you payAmazon US: current deals, useful picks and tech finds.Check DealsClean PCRecommendedOne scan can reveal what keeps slowing WindowsLook for cleanup and repair opportunities.Run Scan×
Skip to content
VGSources
AI gaming

Why Are Large Language Models So Terrible at Video Games?

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.

Large language models can describe a game without being able to play it reliably. Playing means repeatedly identifying what is happening, choosing a legal and useful action, seeing what changed, and adjusting while keeping the objective in mind. A mistake at any point can derail the next move; complex games add precise movement, timing, exploration, and long-term planning.

Why answering about a game is easier than playing one

A question about a game usually asks for a response from information already available to the model. Playing is an ongoing control problem. The agent must turn observations into a usable account of the current state, select an action through the game’s interface, and use the result to decide what to do next.

  1. Observe: identify relevant objects, positions, status, and the current objective.
  2. Interpret: infer what matters now and how the game state may change.
  3. Act: issue an input the game accepts, with suitable direction, timing, or precision.
  4. Check: inspect the outcome, notice errors, and update the plan.

This loop may repeat many times before a goal is reached. A fluent explanation does not guarantee accurate perception, reliable control, or a plan that survives new information.

Why an AI can misread what is happening on screen

Images are not automatically a useful game state. To act well, an agent needs to distinguish relevant objects, track their positions and status, and understand what the display says about progress or the current goal. If that initial interpretation is wrong, even a sensible-looking action can be wrong for the situation.

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.
#1 Best Overall
Logitech G F310 Wired Gamepad Controller Console - Blue/Black
  • With broad game support, the Logitech Gamepad F310 works with old standbys to today's biggest titles, so it's easy to set up and use with your favorite games.
  • Profiler software allows the gamepad to be programmed to perform keyboard and mouse commands for games without gamepad support.* * Requires software installation.
  • A familiar control layout that doesn't require a learning curve to be able to use, with all the same buttons as on an Xbox 360.
  • The unique floating D-pad rests on four switches-instead of a single pivot point-making it responsive to quick changes in direction.
  • The six-foot cord lets you lean back and play a comfortable distance from your PC monitor.

The ICLR 2026 LMGame-Bench evaluates 13 models across six games with a modular setup that probes perception, memory, and reasoning. Its authors report limitations in visual state extraction, reflection, spatiotemporal reasoning, and long-context reasoning. These are distinct possible failure points: an agent can misunderstand the state, or understand it and still choose poorly.

The ICLR 2025 BALROG benchmark likewise reports that some models did worse when given visual representations of environments. That finding applies to the models and tasks in its evaluation; it does not mean visual input always hurts performance. The result can depend on the model, game, and way the state is presented.

Rank #2
Sale
VOYEE Switch Controllers for Switch/Lite/OLED/PC, Pro Controller
  • RGB Cool Lightning Bolt & Long Battery Life: Switch controller with lightening bolt style and 9-color LED looks really cool; 4 light modes: solid lights, breathing lights, RGB strobe led light and led off; Fully charged: 3-4 hours, Runtime: 10-15 hours
  • Widly Compatible & One-key Pairing/Wake Up: The switch pro controller is compatible with Switch/Lite/OLED/PC Windows 7/8/10 (only wrok for pc under wired connection); 2 pairing ways; Support one key to wake up your switch console
  • Programmable Button & 3 Speeds Turbo: Switch controllers has simplify complex skill operations with M1/M2 key; Support single and multiple mapping; 3 adjustable burst: 5 shots/s, 12 shots/s and 20 shots/s; Programming and Turbo will maximize game play
  • Sensitive Motion Control & 4-Level Nice Vibration: 6-axis gyro sensor help you react quickly, enhance experience in sports games; Buttons and joysticks are responsive, no lag; Dual vibration motors with 4-level feedback: Strong, Medium, Weak and None
  • Great Gift For All People: This cool switch controller will be great gifts for women, men, girl, boy, family and friends; Packing list: 1 X Wireless switch controller, 1 X Type-C cable, 1 X Detailed user manual, 1 X Cool gift box

Why movement and timing are hard to get right

Knowing roughly what should happen is not the same as controlling a character or cursor precisely. An agent may need to track positions, predict the effect of movement, and act at the right moment. A direction that is broadly correct can still miss a target, produce an invalid move, or fail to make progress.

LMGame-Bench identifies spatiotemporal reasoning as a weakness, and SmartPlay includes spatial reasoning among the abilities it tests across six games and up to 20 evaluation settings. Examples on VideoGameBench include difficulty with accurate cursor control in a physics puzzle and goal-oriented navigation in an adventure game. These illustrate different ways that a high-level intention can fail to become effective control.

Free tools Windows power users keep installed

One-click scans. No signup required.

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.
Rank #3
EasySMX X15 Wireless PC & Steam Gaming Controller, Hall Effect, Black
  • Platform Compatibility: This PC controller is designed for Windows PC, Steam, Switch, Android, and iOS. Xbox-style asymmetric stick layout for PC gamers. Three modes cover all your devices. Please check your device compatibility before purchase
  • Three Connection Modes: 2.4G wireless, Bluetooth, wired USB-C. PC gets native XInput/DirectInput. Switch pairs via Bluetooth, no adapter. This gaming PC controller switches devices seamlessly. Stable wireless minimizes random disconnects during gaming
  • Hall Effect Precision: Hall effect joysticks and triggers eliminate stick drift. This gaming controller for PC delivers smooth, responsive input with no dead zones. Built for FPS, racing, and action games. Long-term precision for competitive PC gaming
  • Back Buttons & Battery: Two programmable back buttons map combos and shortcuts. Textured grips with dual vibration. 1000mAh battery delivers up to 20H playtime. RGB can be turned off. A solid PC controller for gaming with custom back buttons
  • ABXY Layout Switch: Press B + Minus + Plus to swap between PC and Switch modes. Features: 1000Hz polling rate, RGB lighting, turbo. Note: designed without mic jack or gyro sensor

Why agents lose track of goals over long play

Longer games demand more than a good next move. An agent may need to remember its objective, track what it has already tried, and connect a small action now to a later payoff. If it forgets why it is exploring or repeats an unproductive approach, local decisions can look plausible while the overall run goes nowhere.

LMGame-Bench reports limits in long-context reasoning. VideoGameBench examples illustrate the practical symptom: GPT-4o loses track of the primary objective after choosing a Pokémon starter, while an agent wanders in Link’s Awakening while searching for a sword. These examples show observed failures in particular setups, not a controlled explanation for why every model loses focus.

Rank #4
MYSTILUCK Wireless Pro Controller for Switch/Switch 2/Lite/OLED/PC
  • Compatible with Wide Range of Consoles: This controller works with consoles such as Switch 2, Switch, Switch Pro, Switch Lite, and Switch OLED. (Please note): The controller's “HOME” button cannot wake up the Switch 2 console and does not have the C button for voice chat functions. However, all other functions are fully usable, including: dual vibration, 6-axis gyroscope, screenshot function, Hall effect buttons, and turbo.
  • Cool and Colorful Lighting Switch Controller Wireless: It features 7 colors of RGB lighting (Red - Orange - Yellow - Green - Cyan - Blue - Violet) and 4 light modes (Dazzle - Monochrome - Monochrome Breathe - Monochrome Breathe Cycle).
  • Hall Effect Technology for Switch Pro Controller: Experience zero drift and unmatched accuracy with our Hall effect joystick switch. Adaptive trigger feedback with adjustable resistance levels lets you feel every action. With <0.1 ms response time and 256 levels of pressure sensitivity, enjoy instant trigger detection in FPS games. 3+ million clicks on the controller mean a long service life.
  • Dual Motor Vibration, Turbo Function and 6 Axis Gyroscope: The switch 2 controller has two vibration motors with three intensity levels—off, low, and high—and provides exceptional haptic feedback to enhance the gaming experience. The controller also offers three adjustable turbo speeds (5-10-15 Hz), which are particularly suitable for first-person shooter games. In addition, it features a 6 axis gyroscope chip for precise motion control. The physical movements of the players are precisely matched to the actions of their game characters.
  • Reliable After-Sales Support You Can Count On: Your satisfaction is our top priority. Should you experience any quality concerns with your gaming controller, simply reach out to us via our customer service email, and we’ll respond promptly. We stand behind our product with a hassle-free replacement policy—ensuring you’re back to gaming without worry, no questions asked.

BALROG reports partial success on easier games and substantial difficulty on harder tasks, which can demand complex interaction, advanced spatial reasoning, long-term planning, and continued exploration of strategies. As a game grows more demanding, an agent has more opportunities for a perception error, a weak decision, or a forgotten objective to compound.

Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Support on Ko-Fi

Why real-time play adds another constraint

Some games keep running while an agent analyzes an observation and generates an action. That creates a systems problem alongside the reasoning problem: the world may change before the input arrives. The VideoGameBench project identifies inference time as a major bottleneck for vision-language models playing games.

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.
Best Value
GameSir G7 SE Wired Controller for Xbox Series X|S, Xbox One & Windows 10/11, Plug and Play Gaming Gamepad with Hall Effect Joysticks/Hall Trigger, 3.5mm Audio Jack (White)
  • Versatile compatibility: supports Xbox Series X/S, Xbox One X/S consoles and PC Win10 and above (including the game platform Steam).
  • Precise control: features Hall joysticks and Hall triggers for a comfortable feeling, long service life and improved game accuracy.
  • Plug and Play Convenience: Wired USB connection (removable) for easy setup and instant play without the need for additional drivers.
  • Customizable experience: Includes 2 custom backbuttons that allow users to eliminate false triggers and improve their gaming experience.
  • Impressive gameplay: Provides a pulsating vibration trigger and an asymmetric vibration grip motor for intense tactile feedback.

VideoGameBench Lite pauses the environment during model inference. Pausing lets researchers examine gameplay without real-time inference pressure; it does not remove challenges such as interpreting the state, controlling accurately, or planning over a long horizon. Latency can worsen real-time performance, but it is not a complete explanation for game-playing failures.

What game benchmarks do—and do not—show

There is no single score that captures “how good an AI is at games.” Evaluations differ in their games, observations, action interfaces, time limits, and support for memory or planning. The benchmark name alone is not enough to establish that two results measure the same capability.

Benchmark Scope and setup described by its source What the evidence helps examine
LMGame-Bench Six games; 13 evaluated models; modular probes for perception, memory, and reasoning. Visual state extraction, reflection, spatiotemporal reasoning, and long-context reasoning.
BALROG Game environments with varying difficulty, evaluated with LLMs and VLMs. How performance differs between easier games and harder tasks involving complex interaction and planning.
VideoGameBench 23 curated video games, with real-time and paused Lite modes. Gameplay examples and the effect of real-time inference pressure versus a paused environment.
SmartPlay Six games and up to 20 evaluation settings. Agent capabilities that include spatial reasoning.
GameBench Nine game environments in a 2024 study focused on strategic reasoning. Strategic performance under the evaluated models, tasks, opponents, and reasoning or planning scaffolds.

In that 2024 GameBench study, none of the evaluated models matched the human baseline. GPT-4 in the tested setup could perform worse than random action; chain-of-thought and reasoning-via-planning scaffolds improved scores but did not reach human performance. These results describe that paper’s particular evaluation, not every current model or every game.

  • Observation: Is the agent looking at pixels, or receiving structured information about the game state?
  • Timing: Does the game continue during inference, or pause while the model responds?
  • Task difficulty: Is it a short, simple challenge or a complex game requiring exploration and long-term planning?
  • Agent support: Is the evaluated system a base model, or does it add prompts, memory, tools, or planning scaffolds?

A score answers a question about a particular setup. Changing any of these conditions can change what the test measures and how well a system performs.

What’s actually slowing this PC down?

Pick the symptom - the matching free tool is one click away.

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.

So, are large language models actually terrible at video games?

“Terrible” depends on the game and the conditions. Benchmark findings show that models can manage some easier tasks yet struggle with harder play involving visual interpretation, precise control, sustained goals, exploration, and planning. Scaffolds can improve results in some evaluations, but improvement in one setup does not establish reliable play across different games or interfaces.

Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.

Leave a Reply

Your email address will not be published. Required fields are marked *

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.

Read next

Recommended PC Tool
Recommended PC Tool
Outdated Drivers Are Slowing You DownFree scan - exact matches
Windows Errors? Fix Them Before They SpreadFree repair scan

Two free Windows tools

One Free Minute Could Fix That PC

Before you go - each of these free tools takes about a minute and tackles what quietly slows a Windows PC down.

Special offer. View Outbyte info, uninstall instructions, EULA, and Privacy Policy.