(Washington, D.C., May 29, 2026) – U.S. Secretary of Agriculture Brooke L. Rollins today announced payment rates and the enrollment period for the Assistance for Specialty Crops Farmers (ASCF) program. The U.S. Department of Agriculture (USDA) will issue $1.625 billion in payments to eligible specialty crop producers in response to elevated input costs and market disruptions resulting from foreign competitors engaging in unfair trade practices that impeded specialty crop exports.

We may earn a commission from links on this page.

If you’ve been eyeing the Garmin Forerunner 970, you may be drawn to two of its buzz-worthy features: running economy and step speed loss. But there’s a catch: You can’t access either metric without purchasing a compatible Garmin accessory like the HRM-600 heart rate monitor, which will run you $170.

Is this extra training insight worth the additional investment? To find out, I’ve been wearing one on my runs for the past several weeks, including a half-marathon race in early May. Here’s what I’ve learned about myself, and what you need to know before buying an accessory to get the most out of your running data.

What is “running economy,” and how does Garmin calculate it?

Running economy technically refers to how much oxygen you consume at a given pace. But when runners talk about it, we’re referring to a basic principle that says the more efficient you are, the less energy you’ll burn to run at the same speed. Garmin translates this into a practical score by analyzing your heart rate, speed, and running dynamics. The key input in their calculation (which requires the heart rate monitor) is something called “step speed loss.”

To provide your score, Garmin says it needs data from at least 5–7 runs. After that, you can find your Running Economy score in the Garmin Connect app by tapping More (bottom right) > Performance Stats > Running Economy.

What is “step speed loss,” and how does Garmin calculate it?

Simply put, “step speed loss” (SSL) is a measure of braking. Garmin defines it as the difference between your forward speed at the moment your foot first contacts the ground and your minimum forward speed during that step’s stance phase. Picture the brief window when your foot is planted and your body is passing over it. It’s measured at the chest via a heart rate monitor, and expressed in centimeters per second. You want that number to be as low as possible.

To understand why SSL matters, it helps to think about what’s happening physically with each stride. A high SSL means your body is somehow hitting the brakes when your foot hits the ground, and you have to do a lot more work just to get back to where you were. A lower SSL means your foot is spending less time dragging against the ground and your stride is closer to a smooth, continuous wheel (rather than going brake-accelerate-brake-accelerate on a loop). 

Several things can drive you SSL up: low cadence, overstriding (landing with your foot too far in front of your hips), or anything that creates a heavy, stomping contact. You can try to bring your SSL down by running with a lighter, shorter stride, with your foot landing right beneath the body (rather than reaching ahead or dragging behind).

One thing I discovered while digging into all of this: some gait metrics, like cadence and ground contact time, seem to improve on their own when you run faster, because speed itself encourages better mechanics. It’s a bit of a “chicken before the egg” scenario, but for proper running form and speed. 

What Garmin’s running dynamics taught me

I’ll spare you all my typical face-saving disclaimers about how I’m an extremely average runner—almost always in the 50th percentile of every race I run, in fact. Let’s take a look at the numbers. In these graphs, each dot represents a percentile band compared to other runners. Purple sits at the top (95th percentile and above), blue covers 70–95%, green is the middle range at 30–69%, orange drops to 5–29%, and red represents the bottom fifth percentile. Garmin says the green, blue, or purple zones are typical for more experienced runners, which tracks for me.

If graphs aren’t your thing, here are the stats from the run listed out in my Garmin Connect app.

My average step speed loss for the race was 8.2 cm/s, translating to an average SSL percentage of 2.85%. That means at any given step, I was losing roughly 2.85% of my forward speed during ground contact. 

A bright spot for me is that there’s no obvious upward creep in the second half of the race, which is exactly where I’d expect to see my form break down due to fatigue. I checked out the green spikes around the 0:49-1:15 mark, and it turns out they correspond to a downhill portion of the race. It makes sense that I was subconsciously breaking my stride instead of blowing out my quads and charging down those hills.

The other numbers that catch my eye are cadence and vertical ratio. My cadence averaged 181 steps per minute, which tickles me, as180 spm has to be by far the most common “optimal running cadence” all runners know. Unfortunately, my vertical ratio isn’t as pretty a picture: Vertical ratio also measures running efficiency, but by dividing vertical oscillation by stride length. At 8.5%, I’m below average in this metric. Instead of digging into that right now, let’s turn the other measure of efficiency I care about today: running economy.

How useful is Garmin’s running economy score?

The Running Economy screen (dated from the race on Saturday, May 16) shows a score of 214, placing me in the “Trained” category. Garmin’s own description of the score is that my running economy is at a good level with room to improve, and meaningful gains will likely require gradual increases in both intensity and volume. Progress at this level may come more slowly, it tells me, but it’s achievable.

For now, I’ll take that at face value. As I’ve previously covered, there are tangible ways to improve this metric, but a lot of it comes down to genetics. What this number does is give me something concrete to track. If the score nudges into green, blue, or purple territory over the coming months, that’s a signal my training is translating into measurable efficiency gains. If it stagnates while volume and intensity go up, that’s worth investigating. Running Economy scores are only as meaningful as the underlying data, so I plan to log more runs with the HRM-600 to build a better picture.

Do you actually need the HRM-600?

Personally, the running economy score is an interesting metric to have, particularly with my newfound understanding of how SSL works. Beyond the single-race snapshot, I think there’s real value in simply wearing the HRM-600 consistently and tracking my personal SSL range over time. 

For most runners, I don’t think SSL is something to obsess over. Percentile comparisons can be fun, but there comes a point where it’s better to tune them out. What matters more is whether your SSL is trending in a useful direction.

Whether or not paying $170 for the HRM-600 is worth it depends on how deep you want to go. Even without obsessing over the number, SSL is a helpful form-check tool for me. When it creeps up on a run, it’s a useful nudge to perform a mental audit to figure out what’s happening: Am I over-striding? Is my foot landing ahead of my hips instead of underneath them? How’s my cadence doing? A few cues I keep coming back to: shorten the stride, let the foot land beneath the body, and think “light and quick” rather than “push and drive.”

If you’re a data-oriented runner who’s already squeezing value out of your Forerunner 970 and looking for the next layer of insight, running economy and step speed loss are natural next steps. If you’re still working on nailing the basics, you’ll get more return from simply running more consistently. Save that money for race registration fees instead.

It’s difficult to get away from AI inside apps and on devices now, and if you’re making a lot of use of these tools, you might be thinking about subscribing to a paid tier. While all the major AI chatbots offer free access, subscription plans offer more features and extended usage limits. Of course, there are a host of options to pick from—several big AI services, all with multiple plans at different price points. This guide should help you make some sense of the various offerings, and help you decide on the right one (or two) for you.

What’s not included here is any assessment of which AI models are “better” than others. That can be difficult to measure, especially as models frequently change. Using the free tiers for these bots should give you a good idea of which AI models you like best, and which give the most relevant responses to your queries.

Gemini

Google doesn’t specify what the “standard limits” are for free Gemini usage, but it does say that AI Plus users get twice those limits, while AI Pro users get four times those limits. If you pay for AI Ultra, you get five times or 20 times the limits of AI Pro, depending on how much you’re spending. The AI Pro and AI Ultra plans give you access to the latest AI models from Google.

Context windows go up with each tier, too, which is how much the AI can keep in its memory per conversation. The standard on free plans is 32K tokens (the blocks of info that AI bots work with), which add up to around 24,000 words. For AI Plus plans, that goes up to 128K tokens (about 96,000 words), and for all the other paid plans, you get one million tokens (about 750,000 words).

You can generate text, code, audio, images, and video with Gemini. It features a deep research mode, the ability to create custom AI bots with Gemini Gems, and scheduled actions. Most features are available on all the plans, but you’ll find some exceptions. Video generation requires a paid plan, as does image editing. The new Gemini Spark AI agent is currently available only to AI Ultra subscribers.

Gemini makes sense if you already use a lot of Google apps (Gmail, Google Maps, Google Docs) and Android. After all, it’s built right in. It’s also worth noting that you get a host of extras included with these AI plans: more Google One storage at each tier, as well as YouTube Premium, Google Home Premium, and the new Google Health Premium (for AI Pro and AI Ultra plans).

• Paid plans: AI Plus ($7.99 a month), AI Pro ($19.99 a month), AI Ultra (either $99.99 or $199.99 a month)

ChatGPT

It’s not particularly easy to figure out usage limits on the ChatGPT plans, but what we know for sure is that the Pro plans give you five times or 20 times the usage of Plus plans, for $100 or $200 a month, respectively. A Pro plan gives you unlimited messages with the latest models too, whereas on Go and Plus, you’re limited to 160 every three hours. Also of note: The Go plan may display ads at times.

For context windows, again, OpenAI doesn’t make it particularly clear. Free users get 16K tokens (roughly 12,000 words), which goes up to 32K tokens (24,000 words) for Plus users, but there’s no mention of the Go plan. Pro users currently get 128K tokens (96,000 words) for GPT‑5.5 Instant, and 400K tokens (300,000 words) for GPT-5.5 Thinking—the best models ChatGPT has at the moment.

ChatGPT is able to generate text, code, and images for you. It has a deep research mode available to everyone, and support for custom AI bots (which it calls GPTs) on the Plus and Pro tiers. In general, paying more gets you more of everything: more messages, more uploads, more image generations, and access to the latest models first.

There’s no wider ChatGPT suite of tools behind the AI chatbot, though it can connect up to a host of apps—including Photoshop, Figma, Spotify, Apple Music, Airtable, and plenty more. It also has its own experimental browser (ChatGPT Atlas), but it’s very platform-agnostic when it comes to what it works best with.

• Paid plans: Go ($8 a month), Plus ($20 a month), Pro ($100 or $200 a month)

Claude

You wouldn’t really describe Claude’s usage plans as straightforward, but obviously, the more money you pay a month, the more usage you get. Exactly how much usage that works out to “is affected by several factors, including the length and complexity of your conversations, the features you use, and which Claude model you’re chatting with,” according to the official documentation.

Context windows are the same on all three paid plan tiers: 200K tokens (about 150,000 words). What the context window is for free users isn’t specified anywhere. Claude does let you pay as you go, to some extent, if you need extra AI usage on top of what you’re already being charged for your subscription.

Coders frequently choose Claude for development, and, of course, it can also do general text output. It can’t do images or video, though it can do basic visualizations (like charts and diagrams), and also offers the Canva-like Claude Design portal that will generate user interfaces, slideshows, and mock-ups for you. There’s also a deep research tool here as well.

Like ChatGPT, and unlike Gemini, Claude isn’t really tied to any particular ecosystem. However, it can connect to a multitude of other apps that you can then run through Claude, including Canva, Gmail, Slack, and Uber. There’s also a Skills feature, where you can create customized instruction prompts for Claude to run again and again.

• Paid plans: Pro ($20 a month), Max ($100 or $200 a month)

Copilot

Copilot is unusual here, because (for individuals, rather than software developers) it’s tied into Microsoft 365 (previously Microsoft Office) subscriptions. If you sign up for Word, Excel, and PowerPoint, then you get Copilot as well, at the two main pricing tiers.

While Microsoft isn’t too precise about Copilot’s usage limits, it does say that Personal subscribers get “higher than free” usage limits, and Premium subscribers get the “highest” limits. This applies to image generation too, while AI agents and audio creation are exclusive to the Premium tier. Microsoft doesn’t say anything about context limits for individual users.

Copilot can create text and images, deploy a deep research mode, and connect to third-party services such as Gmail and Dropbox. There aren’t quite as many of the AI bot features as you’ll find elsewhere, like scheduled tasks or custom-made AIs for specific purposes, though there is a neat Study and Learn mode.

As with Gemini and Google, the primary reason to pick Copilot for your paid AI subscription is if you’re already invested in Windows, the Edge browser, and Microsoft’s office apps. Copilot integrates deeply into these pieces of software, and so is easier to call on if you need AI assistance.

• Paid plans: Personal ($9.99 a month), Premium ($19.99 a month)

Perplexity

Perplexity models itself as more of an AI research tool and search engine, rather than an AI chatbot in the more general sense—like ChatGPT or Gemini. You can actually use Gemini, ChatGPT, and Claude models inside Perplexity, as well as Perplexity’s own models (or use several in combination).

There’s little official information from Perplexity about exact usage limits or context windows, though the more you pay, the more usage you’ll get: The Max tier is described as offering the “highest usage and top performance”, and gives you access to the “most advanced AI reasoning models”.

Perplexity does offer tools like deep research and custom AI silos that it calls Spaces—complete with their own custom prompts and specific models. You can create images and videos with Perplexity, but it relies on third-party models for the task, and (anecdotally at least) usage limits seem to be quite tight.

The primary reason to sign up for a Perplexity subscription is if you find its focus on web search and information gathering useful. It also has its own browser, as well as some agentic tools that can carry out actions online for you. It’s more focused than other AI chatbots, which may or may not be what you’re looking for.

• Paid plans: Pro ($20 a month), Max ($200 a year)