AI Coding Live News — Real-Time 𝕏/Twitter Updates & Breakthroughs

👀0

🔥82

Fireworks Forge Launches: Ultra-fast Inference Preview Hits 1,500 Tokens/s, Half-Price Background Mode, Nexus Desktop for Codex and Claude Desktop, Sovereign Private Preview

At its first Forge conference on November 3, open-model inference platform Fireworks launched a batch of products. Ultra-fast, powered by NVIDIA Blackwell, is in public preview at up to 1,500 tokens per second on DeepSeek V4.1 Flash, cutting a 30-turn agent task from 1 minute to 8 seconds. Background Mode runs deferrable work at half price, alongside multi-day prompt caching and higher rate limits. Nexus Desktop brings open-model routing into the Codex app, Claude Desktop and OpenCode, with Pro plans from $8 per month. Fireworks also announced No Eval, serverless LoRA inference, the Fireworks Sovereign private preview and a 60+ partner Frontier Ecosystem.

Key Takeaways

Ultra-fast public preview: on NVIDIA Blackwell, up to 1,500 tokens per second on DeepSeek V4.1 Flash; in Fireworks' benchmark a 30-turn agent task fell from 1 minute to 8 seconds (Fireworks blog)
Cost options: Background Mode runs deferrable work at half price; multi-day prompt caching cuts the cost of resuming long context; higher serverless rate limits and Elastic Burst are generally available
Nexus Desktop launches: personalized routing across open and closed models inside the Codex app, Claude Desktop and OpenCode; Pro from $8 per month, Teams from $10 per seat per month
Training: No Eval (private preview) turns production traces into task evals; serverless LoRA inference (public preview) starts with GLM 5.3 Flash at per-token pricing with no idle charge; free training on Muse Glimmer 30B through November 30
Enterprise: Fireworks Sovereign (private preview) runs the inference engine in a customer's own cloud or data center; Fireworks says it already manages over 25,000 GPUs in customer environments; plus a 60+ partner Frontier Ecosystem

Read full story

Linked Models & Agents:

⚡ DeepSeek-V4.1-Flash ⚡ GLM-5.3-Flash⚡ Muse Glimmer 30B⚡ Ember-1🤖 Codex 🤖 Claude Desktop 🤖 OpenCode 🌐 Fireworks 🌐 NVIDIA

#Fireworks#Forge#Inference#Ultra-fast#Background Mode#prompt caching#Nexus#Sovereign#open models#LoRA

OpenAI DevelopersOpenAIDevs·Collected 14d ago

👀86

🔥84

View on 𝕏 ↗

OpenAI API adds 12-hour prompt caching: GPT-5.6+ models can keep shared context cached for half a day, with writes at 2.5x input and reads still at 0.1x

On Oct 20 OpenAI added 12-hour prompt caching to the Responses API: GPT-5.6 and later models can set prompt_cache_options.ttl to "12h" so instructions, tool definitions and reference material stay billed at the cached rate even when requests are hours apart. 12-hour writes cost 2.5x the ordinary input price versus 1.25x for the default 30 minutes; read rates are unchanged at 0.1x on most models and 0.05x on GPT-6.1 Sol. One request can mix 12-hour and 30-minute breakpoints.

Key Takeaways

New option: prompt_cache_options.ttl in the Responses API accepts "30m" (default) or "12h", and prompt_cache_breakpoint.ttl overrides it per breakpoint, for GPT-5.6 and later models
Rates (vendor-stated): 30-minute writes at 1.25x and 12-hour writes at 2.5x ordinary input; reads at 0.1x on most models and 0.05x on GPT-6.1 Sol, same for both lifetimes
Our estimate from the documented rates: reusing one prefix 6 times an hour apart costs about 6x1.25=7.5x input with 30-minute caching (rewritten each time) versus about 2.5+5x0.1=3.0x with 12-hour caching
Up to 4 cache writes per request and a 1,024 visible-token minimum prefix; usage now splits writes into ttl_30m_tokens and ttl_12h_tokens
"12h" is a minimum lifetime, not a deletion deadline, and does not guarantee a hit or lower latency; GPT-5.5 and earlier keep using prompt_cache_retention

Read full story

Linked Models & Agents:

⚡ GPT-6.1 Sol⚡ GPT-5.6🌐 OpenAI

#Prompt Caching#Responses API#Agent#OpenAI

OpenAI Developers@OpenAIDevs·Collected 42d ago

👀65

🔥90

View on 𝕏 ↗

OpenAI improves GPT-6 prompt caching: higher default hit rates, up to 90% off cached input

OpenAI Developers (@OpenAIDevs) announced prompt-caching improvements for the GPT-6 API: higher default cache-hit rates so more input tokens get cached-input discounts of up to 90%, helping agents run faster and cheaper. OpenAI docs also point developers to the Prompt Caching Dashboard and miss Diagnostics.

Key Takeaways

GPT-6 API prompt caching now hits more often by default for agent workloads.
Cached input tokens discounted up to about 90%.
Dashboard monitors hit rates; Diagnostics help fix cache misses.

Read full story

Linked Models & Agents:

⚡ GPT-6 Sol ⚡ GPT-6 Luna ⚡ GPT-6 Astra 🤖 openai-codex 🌐 OpenAI

#GPT-6#Prompt Caching#OpenAI#API pricing#Agents

@OpenAI@OpenAI·Collected 42d ago

👀47

🔥88

View on 𝕏 ↗

OpenAI launches GPT-6 Sol & Luna: dual enterprise workhorses cut API pricing by 50% down to $0.10/M

OpenAI introduced GPT-6 Sol and GPT-6 Luna, faster and more affordable models that carry forward much of GPT-6 Astra’s strength for work at scale. Caching and inference efficiency cuts yield 50% lower API prices versus GPT-5.6 promotional pricing: Sol is priced at $2/$10 per million tokens for deep reasoning, while high-volume Luna costs just $0.10/$0.50 per million tokens with a 90% prompt caching discount.

Key Takeaways

Tiered dual-model architecture: Sol for deep reasoning and complex coding agents, Luna for ultra-high-throughput workflows.
50% price slash: Sol at $2/$10 per million tokens; Luna down to $0.10/$0.50 with 90% prompt caching discount.
Available immediately in OpenAI API (gpt-6-sol, gpt-6-luna), ChatGPT Work, and Codex.

Read full story

Linked Models & Agents:

#AI#LLM#Coding

Kimi.ai@Kimi_Moonshot·Collected 43d ago

👀60

🔥91

View on 𝕏 ↗

Moonshot: Kimi K3 now on Amazon Bedrock with coding/agent workflows and explicit prompt caching

@Kimi_Moonshot announced Kimi K3 is now on Amazon Bedrock for coding, document analysis, and extended agent workflows, with Bedrock access, encryption, and auditing controls, plus explicit prompt caching. AWS published a Bedrock model card so teams can call K3 inside the AWS stack.

Key Takeaways

Kimi K3 is live on Amazon Bedrock.
Targets coding, document analysis, and extended agent workflows with Bedrock access/encryption/auditing.
Explicit prompt caching supported; start via the AWS Bedrock model card.

Read full story

Linked Models & Agents:

⚡ Kimi K3 🌐 Moonshot AI (月之暗面)🌐 Amazon AWS (Nova)

#Kimi K3#Amazon Bedrock#Moonshot#prompt caching#coding agents#AWS

vechen@miu21590·Collected 44d ago

👀84

🔥88

View on 𝕏 ↗

Mid-task Jev in Codex raises/lowers GPT-6 Astra reasoning effort; author reports ~50% Astra cost cut without breaking prompt cache

@miu21590 demoed using Jev inside a running Codex session to raise/lower GPT-6 Astra reasoning effort—more thinking when stuck, less on routine steps—reporting ~50% lower Astra costs and faster runs without breaking prompt caching. Versus the usual pre-task Jev model pick, this wires the decision model mid-run for cost control.

Key Takeaways

Call Jev mid-Codex-run to raise/lower GPT-6 Astra reasoning effort.
Author tests: ~50% lower Astra cost, faster, prompt cache intact.
Moves Jev from pre-task model pick to in-task cost control.

Read full story

Linked Models & Agents:

🤖 codex 🌐 OpenAI

#Codex#Jev#GPT-6#Astra#reasoning effort#prompt cache#Cost

Google for Developers@GoogleDevs·Collected 46d ago

👀88

🔥95

Google AI Studio Unveils Managed Agents Architecture with Antigravity Preview Model

Google AI Studio introduced its Managed Agents architecture powered by the antigravity-preview-09-2026 model. The update features optimized prompt caching for software repositories and high-throughput File APIs for frictionless multi-turn agent interaction.

Key Takeaways

Deploys Managed Agents architecture with specialized antigravity-preview model
Repo-level prompt caching slashes repetitive codebase context costs by 80%
High-throughput File API allows seamless mounting of large multi-file projects

Read full story

Linked Models & Agents:

⚡ Antigravity Preview 09.2026🤖 Managed Agents 🌐 Google

#Google#AI Studio#Managed Agents#Antigravity#Coding AI

Dana@MatureIntellect·Collected 58d ago

👀20

🔥85

View on 𝕏 ↗

Bench v2.1.0: constitution + three-stage edit review for Claude Code, installer/CLI, 1036 tests

Dana (@MatureIntellect) shipped Bench v2.1.0: Claude Code runs under a written constitution with three-stage edit review and a tamper-evident ledger. v2.1 adds prompt caching, per-stage model control, installer/CLI, concurrency locks, weekly cost/latency reports, and 1,036 passing tests.

Key Takeaways

Release: Bench v2.1.0 (benchgov.com).
Governance: constitution + three-stage edit review + tamper-evident ledger.
Eng: installer/CLI, prompt caching, concurrency locks, 1036 tests, cost/latency reports.

Read full story

Linked Models & Agents:

🤖 Claude Code (Anthropic)

#Bench#v2.1.0#Claude Code#Governance#prompt caching#agent review

DeepSeek@deepseek_ai·Collected 65d ago

👀88

🔥93

DeepSeek unveils API Prompt Caching rates dropping cached input pricing to ¥0.1 per 1M tokens

DeepSeek introduced automatic API Prompt Caching across its developer platform. When re-sending system prompts, large repository contexts, or multi-turn chat history, cached input tokens are billed at just ¥0.1 per million tokens (~$0.014/M), reducing operating expenses for autonomous coding agents by up to 85%.

Key Takeaways

Cached input tokens cost only ¥0.1 (~$0.014) per million tokens with automated 64k prefix reuse.
Zero manual cache key configuration: gateway automatically handles prefix hash caching.
Dramatically lowers API costs for terminal agents (Aider, Cline, OpenHands) loading full repos.

Read full story

Linked Models & Agents:

⚡ DeepSeek-V4 (Next-Gen MoE)⚡ DeepSeek-R1 (Open Reasoning)🌐 DeepSeek

#Deepseek#Prompt caching#API pricing#Cost reduction#LLM tokens

Anthropic@AnthropicAI·Collected 65d ago

👀43

🔥90

Anthropic upgrades Prompt Caching TTL auto-extension, reducing compute cost for multi-turn agent sessions by 40%

Anthropic enhanced its API Prompt Caching with automated TTL sliding renewal. For sustained agent interactions, active context caches are automatically extended without re-write penalties, reducing overall token write costs by 40% in multi-turn software workflows while maintaining sub-300ms TTFT.

Key Takeaways

Active context caches automatically slide TTL window, avoiding recurring cache write surcharges.
Reduces multi-turn developer session token expenses by an average of 40%.
Maintains sub-300ms time-to-first-token (TTFT) during dense MCP tool calls.

Read full story

Linked Models & Agents:

⚡ Claude 3.7 Sonnet (Hybrid Reasoning)🌐 Anthropic

#Anthropic#Prompt caching#Claude 3 7#API pricing#Developer experience