GPT-6.2 Astra Scores 174 on Epoch Index, Matches Fable 5.2
OpenAI's GPT-6.2 Astra model has achieved a score of 174 on the Epoch Capabilities Index, placing it on par with Anthropic's Fable 5.2 and exceeding pre-launch reasoning trends. While OpenAI claims superior performance on the 'Agents' Last Exam' at 5% of the competitor's cost, independent benchmarks from Artificial Analysis show Astra scoring 62 against Fable 5.2's 64. The model also scored 62.1 on the AA Intelligence Index, falling short of market expectations for its Sol variant, which received 57.3. Despite these mixed results, Astra is highlighted for significant cost efficiency compared to previous iterations.
Editorial 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
GPT-6.2 Astra Scores 174 on Epoch Capabilities Index, Matching Fable 5.2
According to testing conducted by EpochAIResearch prior to launch, the AI model GPT-6.2 Astra achieved a score of 174 on the Epoch Capabilities Index (ECI). This performance places it on par with the Fable 5.2 model and significantly exceeds the trajectory predicted by pre-Astra trends for reasoning models. The analysis relies exclusively on internal benchmarks used before the official release. The source notes that the ECI is a dynamic metric that will be updated as additional benchmark results become available, meaning these specific values are subject to future revision. This announcement highlights a notable shift in the capabilities landscape for advanced reasoning models.
Read sourceOpenAI Claims GPT-6.2 Astra Outperforms Claude Fable 5.2 on Agents' Exam
OpenAI has announced that its new model, GPT-6.2 Astra, achieves a score 5.5 points higher than Anthropic's Claude Fable 5.2 on the 'Agents' Last Exam.' The company states this performance is delivered at just 5% of the cost per task compared to the competitor. Additionally, GPT-6.2 Astra is reported to be 20% cheaper than OpenAI's previous iteration, GPT-6.1 Astra. However, the source notes a limitation: while efficient and comparable to Fable 5.2 in general metrics, GPT-6.2 Astra falls short in the 'Frontier Code 2.0' benchmark when directly compared to Fable 5.2. The analysis suggests the models are currently more on par with each other rather than GPT-6.2 clearly outperforming Fable across all domains.
Read sourceGPT-6.2 Astra Scores 62.1, GPT-6.2 Sol Scores 57.3 on AA Intelligence Index
A recent post from user pankajkumar_dev reports specific performance metrics for two AI models on the AA Intelligence Index. The model GPT-6.2 Astra achieved a score of 62.1, while GPT-6.2 Sol scored 57.3. The source text includes a summary statement asserting that Anthropic is better than these OpenAI models based on this comparison. The post links to an external source for further details but does not provide the underlying data or methodology within the text itself. This represents a direct reporting of benchmark results and a comparative opinion regarding AI model capabilities.
Show 3 older updatesHide older updates
GPT-6.2 Astra Ranks #62 on Intelligence Index with 57 SOL Score
A social media post from user Mr_Salio reports that the AI model GPT-6.2 Astra has achieved a ranking of #62 on an unspecified Intelligence Index. The post further details that the model received a score of 57 for its 'sol' metric, which is associated with a value of 6.2. The source text includes a link to the full data or context but does not provide additional details regarding the methodology of the index, the organization responsible for the ranking, or the specific date of the evaluation. The information is presented as a direct factual claim within the atomic post.
Read sourceAstra 6.2 Ranks Second on Intelligence Index with High Efficiency
The AI model Astra 6.2 has achieved a score of 62 on Artificial Analysis' overall Intelligence Index, placing it as the second-best model in the ranking. It trails only Fable 5.2, which scored 64. Despite the slight difference in intelligence scores, Astra 6.2 is highlighted for its exceptional efficiency. The source notes that the model reaches performance parity with Fable 5.2 on various benchmarks while operating at a fraction of the cost associated with even more expensive models like Opus 5.5. This performance underscores Astra 6.2's capability to deliver high-level results with optimized resource usage.
Read sourceGPT-6.2 Astra and Sol Score Below Expectations on AA Intelligence Index
According to a post by user pankajkumar_dev, the newly released GPT-6.2 Astra model achieved a score of 62.1 on the AA Intelligence Index, while its counterpart, GPT-6.2 Sol, scored 57.3. The analysis indicates that these results represent a smaller performance jump than anticipated for both models. Specifically, GPT-6.2 Astra remains behind the Fable 5.2 model in rankings. Additionally, the GPT-6.2 Sol model was expected to achieve a significantly higher score than the 57.3 it ultimately received. The source characterizes the outcome as underwhelming relative to market or community expectations.
Read source