Google's most capable Gemini 2.5 model with a 1 million token context window and deep thinking mode. Google lists it on the Gemini API deprecations schedule with Gemini 3 Pro as the recommended replacement.
Google's most capable thinking model with native multimodal understanding.
This model is still tracked for research and discovery, but it is excluded from default public rankings until it returns to active status.
---
Quality Score
1474
Arena ELO
Undisclosed
Parameters
1M
Context
Sign in to join the discussion
0
Downloads
0
Likes
Mar 2025
Released
Launches
3
Benchmarks
19
Research
7
General
1
Recent launch, pricing, benchmark, and API signals linked to this model or its provider.
Google DeepMind 🤝 @A24 We’re launching a research partnership with A24 to ensure the tools of the future are shaped by the creators who use them. Find out more → https://t.co/KN3HdGVjGS https://t.co/IUD7rkcRQS
View source
Quality: 27/100 | Price: $3.438/M tokens | Output: 0 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Arena-Hard-Auto official Gemini-2.5 judged score 79.0 with CI -2.1/1.8
Our Robotics Accelerator has launched with 15 startups helping shape the future of physical AI in Europe. 🤖 This three-month program will connect them with access to our AI stack, Gemini Robotics models and hands-on support from our teams. Meet the companies → https://t.co/HzML8G0YXW
View sourceWhen millions of AI agents interact with each other, new collective behaviors can emerge. 🌐 Together with @schmidtsciences, @coop_ai, @ARIA_research and supported by @GoogleOrg, we’re launching a $10M research fund to help understand how AI systems behave as a group. → https://t.co/Ddq19nQsa5
View sourceQuality: 27/100 | Price: $3.438/M tokens | Output: 0 tok/s | MMLU: 0.862% | HumanEval: 0.801%
View sourceArena-Hard-Auto official Gemini-2.5 judged score 79.0 with CI -2.1/1.8
View sourceGoogle DeepMind 🤝 @A24 We’re launching a research partnership with A24 to ensure the tools of the future are shaped by the creators who use them. Find out more → https://t.co/KN3HdGVjGS https://t.co/IUD7rkcRQS
Our Robotics Accelerator has launched with 15 startups helping shape the future of physical AI in Europe. 🤖 This three-month program will connect them with access to our AI stack, Gemini Robotics models and hands-on support from our teams. Meet the companies → https://t.co/HzML8G0YXW
When millions of AI agents interact with each other, new collective behaviors can emerge. 🌐 Together with @schmidtsciences, @coop_ai, @ARIA_research and supported by @GoogleOrg, we’re launching a $10M research fund to help understand how AI systems behave as a group. → https://t.co/Ddq19nQsa5
In Sierra Leone, a surging student population is outpacing available teachers. Our latest research explores how AI can act as a partner to support educators in these environments – amplifying their reach without replacing their essential expertise and skills. 🧵 https://t.co/8OUO9GamSF
DiffusionGemma is our new experimental open model with up to 4x faster output on dedicated GPUs. Instead of predicting word-by-word, it generates entire blocks of text simultaneously. This lets the model self-correct and format complex markdown in real time. https://t.co/S62OSbfWff
Quality: 27/100 | Price: $3.438/M tokens | Output: 0 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Quality: 27/100 | Price: $3.438/M tokens | Output: 120.586 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Quality: 27/100 | Price: $3.438/M tokens | Output: 131.958 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Quality: 27/100 | Price: $3.438/M tokens | Output: 140.849 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Quality: 34.6/100 | Price: $3.438/M tokens | Output: 142.149 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Quality: 34.6/100 | Price: $3.438/M tokens | Output: 148.078 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Quality: 34.6/100 | Price: $3.438/M tokens | Output: 134.31 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Quality: 34.6/100 | Price: $3.438/M tokens | Output: 125.641 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Quality: 34.6/100 | Price: $3.438/M tokens | Output: 131.75 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Quality: 34.6/100 | Price: $3.438/M tokens | Output: 134.632 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Quality: 34.6/100 | Price: $3.438/M tokens | Output: 139.75 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Quality: 34.6/100 | Price: $3.438/M tokens | Output: 139.046 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Quality: 34.6/100 | Price: $3.438/M tokens | Output: 137.165 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Quality: 34.6/100 | Price: $3.438/M tokens | Output: 125.219 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Quality: 34.6/100 | Price: $3.438/M tokens | Output: 125.219 tok/s | MMLU: 0.862% | HumanEval: 0.801%
Quality: 34.6/100 | Price: $3.438/M tokens | Output: 125.219 tok/s | MMLU: 0.862% | HumanEval: 0.801%
The Frechet Inception Distance (FID) is the de facto arbiter of image generation, yet most papers report just a single number from a single trained model using a single sampling seed. How reproducible is that number if we retrain the model, or merely resample from it? In this paper, we treat FID as a random variable on a two-axis panel of training and generation seeds, and measure its variance directly on several hundred SiT networks trained on class-conditional ImageNet 256x256. We report surprising findings: (a) Retraining the model using the same recipe with a different seed moves FID 3.2x more (in Inception feature space) than redrawing samples from a fixed network. (b) That gap is driven by three factors: random initialisation, data ordering, and the per-step Gaussian noise of the flow-matching loss. (c) Increasing compute or model size barely tightens the spread, holding the FID coefficient of variation (CoV) inside a 1-2% band. (d) Per-cell classifier-free-guidance tuning halves the spread but reshuffles which seeds work best, and a lucky training seed reaches the same FID with up to 2x less compute than an unlucky one. Based on these findings, we recommend a new FID evaluation protocol: evaluate under per-cell optimal guidance, treat any FID gap below the empirically measured ~1.3% CoV as inconclusive, and report an error bar over several training seeds rather than a single FID number.
Video generative models ( VGMs) have become a new frontier that can be used not just for video generation but for a multitude of downstream tasks, including world modeling. To advance these tasks, a good video model must understand the physical reality of the world. Evaluating this understanding is an emerging field and has led to the Physics-IQ benchmark, which quantifies this explicitly by comparing model-generated videos to real-world videos of physical experiments. In this work, we present a systematic audit of the Physics-IQ benchmark, expose shortcomings and propose three solutions that sharpen how we can measure physical understanding of VGMs. Specifically, we improve prompt and ground-truth quality to reduce the influence of confounding factors and further introduce a sample-level scoring system that weights each sample and metric equally. Our resulting benchmark, Physics-IQ Verified, refines 57.6\% of all samples and improves over 34.8\% of prompts. In a comparison study using six image-to-video generative models, we observe moderate but meaningful ranking changes (Kendall's τ= 0.46). We hope Physics-IQ Verified advances the community by providing a more reliable signal toward physically accurate VGMs. The code for the benchmark can be accessed at https://github.com/google-deepmind/physics-iq-benchmark
Reinforcement learning pipelines for Large Language Model (LLM) training often rely on manually redesigned environments between stages, requiring practitioners to heuristically infer which configuration will best improve the current policy. To automate this process, we propose the LLM-as-Environment-Engineer framework in which the current policy model analyzes failure trajectories together with contextual information and proposes modifications to the next-stage training environment configuration. We also introduce MAPF-FrozenLake, a controllable testbed whose generator exposes multi-dimensional environment configurations, making it suitable for studying and benchmarking environment redesign. On this testbed, we condition the environment engineer on structured summaries of policy behavior, failure cases, and environment statistics, from which it produces the configuration for the next training stage. With Qwen3-4B as the backbone, our framework achieves the strongest aggregate performance on our benchmarks, outperforming larger proprietary LLMs (e.g., GPT, Gemini) and fixed-environment training baselines. We further analyze which forms of context are most effective, finding that successful environment updates rely on failure evidence and preserve configurations that already work. Interestingly, the current RL checkpoint serves as a better environment engineer than the original base model, suggesting that policy learning improves the model's ability to diagnose its remaining weaknesses.
Pixel-space diffusion models are trained on full-bandwidth noisy images, yet the useful signal available to the denoiser is strongly frequency dependent. Under rectified-flow diffusion and natural-image power-law spectra, the per-band data-to-noise contour k^{*}(t) = (1-t)^{-2/α} separates a signal-bearing low-frequency region from a noise-dominated high-frequency region at each time t. We show that this implicit coarse-to-fine structure is not merely descriptive: it induces a capacity-allocation problem. A standard pixel-space denoiser must discover the moving bandwidth boundary internally and can spend computation on frequency-time regions where the optimal prediction collapses to deterministic baselines rather than data-distribution modeling. To make this boundary explicit, we introduce Spectral Forcing, a parameter-free, time-conditional 2D-DCT low-pass operator applied to the noisy input before the patch embedder. Its cutoff expands monotonically with the diffusion time and becomes the identity at the data endpoint. Through controlled synthetic experiments, we identify the regime in which the operator is beneficial: coarse patch tokenization and data whose high-frequency content is predominantly noise rather than essential signal. On ImageNet-256 with JiT-700M/32, Spectral Forcing consistently improves both FID and Inception Score across different training epochs, demonstrating robust gains throughout training; at finer tokenization, the spectral forcing is still competitive. We further insert the unchanged operator into SenseNova-U1, a unified text-to-image model, where it improves DPG-Bench and GenEval, showing that the input-side spectral prior transfers beyond class-conditional generation. These results suggest a route to capacity-efficient pixel-space diffusion by showing the signal and hiding the noise.
Many moments in the real world do not wait for a user to ask. A fire starts on a security monitor, an expression flickers across a video call, or a product a viewer wants flashes by in a livestream. Yet today's large models remain mostly turn-based by design: they answer only when addressed, and even video-call apps that appear interactive still operate as question-answer systems, reacting only when polled or prompted. We argue for a different paradigm: a model that is present in the world like a person. It continuously watches what is happening now, decides on its own whether to speak or stay silent, interacts in real time, and delegates to a background model when the problem is hard. To advance interaction models and their adoption across domains, we make two fully open-sourced contributions. First, we release JoyAI-VL-Interaction, an 8B-scale, vision-first VL-interaction model. The model makes the response decision internally, choosing each second to stay silent, respond, or delegate to a background model, and it excels at vision-triggered responsiveness and time awareness. We pair it with a transferable training recipe, from which capabilities we never trained for emerge, such as guiding a shopper through changing app screens or improvising a lecture from a slide deck. Second, we release a complete, deployable system built around that model. The system streams any ongoing video into the model, making it genuinely present in the world. All other components are pluggable, including ASR/TTS modules, memory, visualization UI, and a background brain that can connect to any API or agent. Across six real-world scenarios, human raters prefer JoyAI-VL-Interaction over the in-app video-call assistants of Doubao and Gemini by a wide margin. To our knowledge, this is the first open, vision-driven interaction model released together with its training recipe, data, and complete deployable system.
Temporal Grounding (TG) aims to localize video segments corresponding to a textual query. Prior research predominantly focuses on single-segment retrieval. Real-world scenarios, however, often require localizing multiple disjoint segments for a single query -- a setting we term One-to-Many Temporal Grounding (OMTG). Previous state-of-the-art MLLMs, optimized for one-to-one settings, struggle in this context, often yielding near-zero scores due to a lack of event cardinality perception. To bridge this gap, we present a systematic solution with three key contributions. First, we establish the first comprehensive OMTG benchmark, introducing Count Accuracy (C-Acc) and Effective Temporal F1 (EtF1) as evaluation metrics. Second, we curate a high-quality OMTG dataset comprising 56k samples through a sophisticated construction pipeline. Third, we develop novel temporal and caption reward functions specifically designed for OMTG. In particular, the caption reward leverages Chain-of-Thought reasoning over dense video captions to explicitly guide policy optimization toward both preciseness and completeness. Extensive experiments show our model achieves a new state-of-the-art EtF1 of 43.65\% on OMTG Bench, outperforming Gemini 2.5 Pro and Seed-1.8 by 15.85\% and 15.61\%, respectively.
SWE-Bench Verified resolved rate 53.6