Google DeepMind launches Gemini 4 Argon, matching GPT-6 Astra at one-fifth token cost
On October 1, 2026, Google DeepMind released Gemini 4 Argon, its first non-Flash proprietary model in seven months. The model matches the performance of OpenAI's GPT-6 Astra on several benchmarks while costing one-fifth the token price. It features a 1 million-token output window, nearly 700% larger than rivals, and leads on 13 of 19 benchmarks. The model is rolling out to paid API customers and Google AI Ultra subscribers.
Reference imageEditorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page reads the event directly, while its address stays stable when the title changes.
- Summary covers the current reports
Cross-source coverage
Reporting timeline
Google cancels Gemini 3.5 Pro, releases Gemini 4 Argon to compete with Opus 5.5 and GPT-6 Astra
According to a post by X user Mr_Salio, Google has confirmed it is no longer planning to release Gemini 3.5 Pro, which was originally promised for June. Instead, Google has jumped straight to releasing Gemini 4 Argon, which is now available but not publicly. Argon is being positioned against competing models Opus 5.5 and GPT-6 Astra. Google describes Argon as its most capable model yet. Early results reportedly show Argon beating Astra and Opus on several benchmarks, while trailing on some others. The post also notes that the user had previously stated months ago that Gemini 3.5 Pro would be cancelled.
Google DeepMind Launches Gemini 4 Argon, Matching GPT-6 Astra at One-Fifth Token Cost
On October 1, 2026, Google DeepMind released Gemini 4 Argon, its first non-Flash proprietary model in seven months. The article reports that the new model has surged back into the top three of AI model rankings, matching the performance of OpenAI's GPT-6 Astra while costing only one-fifth as much per token. The launch marks a significant competitive move by Google in the AI arms race, following a period of relative quiet. The source, Chinese tech media outlet 新智元, frames this as a major comeback for Google, emphasizing the cost-performance advantage over its rival. The publication date is October 2, 2026.
Google DeepMind Releases Gemini 4 Argon, Matching GPT-6 Astra at Lower Cost
On October 1, 2026, Google DeepMind released Gemini 4 Argon, its first non-Flash proprietary model in seven months. According to the report from Chinese media outlet 新智元, the model has surged back into the top three of AI model rankings, matching the performance of OpenAI's GPT-6 Astra while being priced at only one-fifth the token cost. The article frames this as a significant competitive move by Google after a period of silence, suggesting the company is reasserting its position in the AI arms race. The summary reflects the source's attribution of performance parity and cost advantage to the new model, based on unspecified benchmarks or comparisons.
Read sourceShow 13 older updatesHide older updates
Google DeepMind Releases Gemini 4 Argon, Matching GPT-6 Astra at One-Fifth Token Price
On October 1, 2026, Google DeepMind released Gemini 4 Argon, its first non-Flash proprietary model in seven months. According to the report from Chinese tech media outlet 新智元, the model has surged back into the top three in AI model rankings, matching the performance of OpenAI's GPT-6 Astra while costing only one-fifth the token price. The release marks a significant competitive move by Google in the AI large language model space, potentially reshaping pricing dynamics and intensifying the race among leading AI developers. The article attributes the performance and pricing claims to unnamed benchmarks and market analysis, suggesting a strategic push to undercut rivals on cost while maintaining high capability.
Google reveals Gemini 4 Argon with 1 million-token output window, 700% larger than rivals
Google has announced its Gemini 4 Argon model, featuring a 1 million-token output window. This capacity is nearly 700% larger than the leading frontier models from competitors OpenAI and Anthropic, according to the company's reveal. The announcement positions Google's latest AI model as a significant advancement in output capacity, potentially enabling more extensive and complex text generation tasks compared to existing models in the market.
Read sourceGoogle reuses internal codename for Gemini 4 Argon public release, ups output limit to 1M tokens
A post on X by user Lentils80 claims that Google has reused the internal codename 'Gemini 4 Argon' for the public release of the model. The user, who asserts their prior information was accurate, notes that the main change since the internal version is an increase in the output limit from 256,000 tokens to 1 million tokens. The post includes a link to further information. This development indicates a significant upgrade in the model's capacity for generating long-form content.
Gemini 4 argon ranks alongside Astra and Opus 5.5 at half the price, user claims
A user on X claims that Google's Gemini 4 argon model now ranks alongside Astra and Opus 5.5 on intelligence benchmarks while costing roughly half the price. The user also highlights Google's generous usage limits, stating that for $20 per month, users can get sufficient usage without worrying about hitting limits. The post expresses jealousy towards Gemini 4 argon users, suggesting it is a competitive option for users of Codex and Claude. The claims are attributed to the user Mr_Salio and are based on unspecified intelligence rankings.
This is a big deal. Google's new Gemini 4 Argon can write up to 1M tokens in a single response, nearly 8x the 128K cap on GPT-6 Astra, Opus 5.5 and Fable 5.1. https://t.co/dbaGCxQNmL
In a post on X, AI analyst Rohan Paul announced that Google's new AI model, Gemini 4 Argon, can generate up to 1 million tokens in a single response. This capability is nearly eight times the 128,000-token limit of competing models including GPT-6 Astra, Opus 5.5, and Fable 5.1. The claim, if accurate, positions Gemini 4 Argon as a significant advancement in large language model output capacity, potentially enabling much longer and more complex generated content than current alternatives. The post attributes this information to an unspecified source linked in the tweet.
Read sourceGemini 4 Argon matches GPT-6 Astra on Artificial Analysis Intelligence Index
A post on X by user kimmonismus reports that Google's Gemini 4 Argon model has achieved a score of 53 on the Artificial Analysis Intelligence Index (AAI), matching the score of OpenAI's GPT-6 Astra and surpassing GPT-6.1 Sol's score of 52. The post also claims Gemini 4 Argon takes the number one spot on the AutomationBench-AA benchmark. Additionally, on the AA-Omniscience benchmark, the author states that Argon hallucinates far less than Astra, with a hallucination rate of 15% compared to Astra's 51%. However, the post notes that Argon's answer accuracy is lower at 50% versus Astra's 63%, and that Argon is more willing to admit uncertainty. The author concludes that the release looks very good and congratulates the Gemini team.
Read sourceGoogle's Gemini 4 Argon outscores GPT-6 Astra and Claude Opus 5.5 on most benchmarks
A post from analyst rohanpaul_ai reports a massive reveal from Google regarding its new flagship AI model, Gemini 4 Argon. The model is claimed to outperform OpenAI's GPT-6 Astra and Anthropic's Claude Opus 5.5 on most industry benchmarks. Its widest lead is in legal work, scoring 19.6% on Harvey's Legal Agent Benchmark compared to 6.7% for Anthropic's Claude Fable 5.1. The model's output limit has been increased from 64K to 1M tokens, described as an industry-leading ceiling. Access is currently limited to three groups: Google's own staff, vetted cyber defenders including government agencies and security companies, and trusted testers providing feedback. Internally, Google reports that Argon agents have freed over 300 TiB of data-center memory, with estimated total savings of 500 TiB to 1 PiB. Additionally, the model made a Rust port of the libgav1 video decoder 2.7x faster by replacing 32,000 lines of SIMD code.
Read sourceGemini 4 Argon tops benchmark with 68.90% score, undercuts Sonnet 5.5 and Opus 5.5 on cost
In a post on X, user Mr_Salio reports that Gemini 4 Argon has achieved the top spot on an unspecified benchmark, scoring 68.90% overall. The post claims that this performance surpasses competitors while also costing less per test than Sonnet 5.5 and Opus 5.5. The author characterizes the combination of strong performance and lower testing cost as a 'pretty crazy combo.' The post includes a link to an external source, but no further details about the benchmark, methodology, or the exact nature of the cost comparison are provided in the text. The claim is attributed to the poster and is presented as a factual announcement of a benchmark result.
Google's Gemini 4 Argon matches GPT-6 Astra on AI Index at 60% of the cost per task
Google DeepMind's new Gemini 4 Argon model has achieved a score of 53 on the Artificial Analysis Intelligence Index, matching OpenAI's GPT-6 Astra (max) and surpassing GPT-6.1 Sol (max) by one point. At its current 50% promotional discount, Gemini 4 Argon costs $1.99 per Intelligence Index task, which is 60% of the cost of GPT-6 Astra (max). The model also boasts the lowest hallucination rate (15%) among leading models, compared to 51% for GPT-6 Astra and 54% for GPT-6.1 Sol. It shows strong agentic performance, ranking first on AutomationBench-AA at 77.5%. The model features a 1M token context window, multimodal input (text, image, video, speech), and text output. Standard pricing is $4/$20 per 1M input/output tokens, currently discounted 50% to $2/$10. The promotion end date has not been confirmed by Google. Gemini 4 Argon is currently being rolled out to selected users and is not publicly available.
Read sourceGoogle releases Gemini 4 Argon, topping 14 of 19 benchmarks against GPT-6 and Claude models
Google has released Gemini 4 Argon, a new AI model that achieves top or tied scores on 14 of 19 benchmarks in Google's internal comparison table. The model is evaluated against OpenAI's GPT-6 Astra, Anthropic's Claude Opus 5.5, and Claude Fable 5.1. Specific results include 77.9% on DeepSWE v1.1 (versus Opus 5.5 at 74.2% and Astra at 74.1%), 51.3% on AutomationBench (versus 42.5%), 84.2% on long context tasks at 256K–1M tokens (versus 71.8%), and 91.7% on LVBench (versus 87.5%). The source, DanDr1s, notes the benchmarks are 'insanely good' but cautions that real-world outputs must be verified, expressing hope the model is not another 'benchmax' that overfits to benchmarks without generalizing.
Read sourceGemini 4 achieves new state-of-the-art on DeepSWE v1.1 with 77.9%, topping multiple benchmarks
A post on X by user Mr_Salio reports that Google's Gemini 4 model, referred to as 'Gemini 4 argon,' has achieved top scores on several AI benchmarks, surpassing competitors including GPT-6 Astra, Fable 5.1, and Opus 5.5. The claimed results include a new state-of-the-art (SOTA) score of 77.9% on DeepSWE v1.1, a #1 ranking on the Vals Index, a 51.3% score on AutomationBench (also #1), a new SOTA of 91.7% on LVBench, and a tied #1 score of 68% on CWE-bench v1. The author describes the performance as a 'terrifying' and 'unexpected comeback,' expressing surprise at the model's benchmark results.
Google releases Gemini 4 Argon, leading 13 of 19 benchmarks and tripling legal test score
In a post on X, user ChrisGPT announced that Google has released Gemini 4 Argon, a stable pro model after over a year and a half of waiting. According to the post, Gemini 4 Argon leads outright on 13 of 19 benchmark results and ties with Astra on cybersecurity. The model achieved 84.2% on GraphWalks at 256K-1M tokens, compared to 71.8% for Astra and 66.8% for Opus 5.5, though it remains behind Muse 1.3 on long-context reasoning. On Harvey's legal benchmark, Argon scored 19.6%, nearly three times the next-best model's 6.7%. It also leads all four knowledge-work evaluations, including 51.3% on AutomationBench against 42.5% for Opus and 41.4% for Astra. Astra still leads on FrontierSWE and Terminal-Bench Science, while Opus 5.5 leads Terminal-Bench 4.0 and PostTrainBench. The post argues that Argon's strongest case is its breadth across knowledge work and long-context reasoning.
Google announces Gemini 4 Argon frontier model with 77.9% DeepSWE v1.1 score
Google has announced Gemini 4 Argon, a new frontier model designed for complex workflows in software engineering, enterprise knowledge work (legal and finance), and cyber defense. The model achieved a score of 77.9% on the DeepSWE v1.1 benchmark, establishing a new state-of-the-art (SOTA) result. According to the announcement, Gemini 4 Argon performs better than GPT-6 Astra, Opus 5.5, and Fable 5.1 across many benchmarks. The model will be rolling out soon, starting with paid API customers and Google AI Ultra subscribers.
Read source